Skip to content

SparkPay MCP server

dev.sparkpay/sparkpaysparkpay.devstreamable-httpv1.0.0Scanned

Stripe billing for indie apps: plans, coupons, subscriptions, customers and revenue reporting.

25 tools: 9 can delete or cancel things, 1 do something that can't be undone, 7 change things and 8 only read.

Tools: 9 delete, 1 irreversible, 7 write, 8 read.
  • Reads private data · 1
  • Reads text others write · 25
  • Can send data out · 1
  • 1 finding

Tools listed without sign-in. Remote posture

With Anlyon

9 tools here can delete or cancel things and 1 does something that can't be undone. Govern them.

  • Anlyon lists the tools and adopts none.
  • Each tool you adopt gets approvals and receipts.
  • Your agent never holds the key.
Govern this server →

A tool your agent calls directly is not routed through Anlyon and is not governed by it. Anlyon Radar only lists and labels, and these labels are the same either way.

What the SparkPay MCP server can do

The SparkPay MCP server is published by sparkpay.dev, a domain the MCP Registry verified. It has 25 tools. It can delete or cancel things (9 tools: delete_coupon, bulk_delete_coupon, revoke_coupon_invite, create_app_kit and 5 more), do things that cannot be undone, such as sending, paying or publishing (invite_to_coupon) and create or change data (7 tools: revenue_summary, revenue_by_app, create_coupon, bulk_create_coupon and 3 more). 8 tools only read.

1 tool reads private data (get_customer); 25 tools bring in text other people can write (list_apps, list_subscriptions, get_customer and 22 more); 1 tool can send data out of your system (invite_to_coupon). On its own it has all three parts of a prompt-injection exfiltration path: private data, text others write, and a way out.

The fixed checks matched 1 finding (annotation-conflict). Each one comes with the text it matched.

Its remote endpoint lists its tools without sign-in, and its tools changed 2 times since Anlyon Radar first saw them.

Last scanned Nov 15, 2026, version 1.0.0.

Tools

Grouped by what each tool does, most consequential first. Open one for its reasons and full description; descriptions are the publisher's own words.

25 tools

Deletes or cancels 9Claude Code: deny
  • archive_planRemove a tier from an app's plan catalog.Reads text others writeNo findings

    Why it is in this group

    • The name "archive_plan" says "archive": a word for removing or ending something.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Remove a tier from an app's plan catalog. WRITE, human-gated: a real write needs an elicitation-capable client and the human must approve; dry_run previews anywhere. Every other plan tool edits tiers that already exist, so a catalog could grow but never shrink, and collapsing a multi-tier page down to a single offer was a dashboard-only job. Stripe products and prices are deliberately left intact, exactly as update_plan_pricing leaves the old price behind: removing the row stops the plan being offered and never reaches into anyone's existing subscription. Refuses while the tier still has non-cancelled subscribers (matched on its Stripe price ids, since subscriptions carry a price rather than a tier slug), and refuses to remove an app's last remaining tier. Run `bun run plans:sync` in the app repo afterwards to regenerate plans.json.

    Annotations the server declares

    • title: Archive Plan
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: true
    • openWorldHint: true
  • bulk_delete_couponDelete the SAME coupon code across many apps in one call: the inverse of bulk_create_coupon, for retiring a portfolio-wide promo.Reads text others writeNo findings

    Why it is in this group

    • The name "bulk_delete_coupon" says "delete": a word for removing or ending something.
    • The input takes a statement or command to run (input.code), which can change anything it can reach.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Delete the SAME coupon code across many apps in one call: the inverse of bulk_create_coupon, for retiring a portfolio-wide promo. WRITE, human-gated: a real delete requires an elicitation-capable client; the human sees every app the code would be removed from (with per-app redemption counts) and must click approve, and confirm_used is only consulted for dry_run previews. Clients without elicitation are refused UNLESS the deployment sets SPARK_PAY_MCP_ALLOW_UNELICITED_WRITES=1 and the call passes confirm_write:true, which is how Claude Code and cron reach these writes; dry_run still works anywhere. An app that never had the code is reported `skipped` rather than failing the batch, since the desired end state already holds there. Returns a per-app outcome matrix plus deleted/skipped/failed counts. Use dry_run:true first to see exactly which apps would lose the code. Pass explicit app_ids (from list_apps) so a delete can't hit an app you didn't name.

    Annotations the server declares

    • title: Bulk Delete Coupon
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: true
    • openWorldHint: true
  • create_app_kitStep 2 of onboarding a new app onto SparkPay.Reads text others writeNo findings

    Why it is in this group

    • The name "create_app_kit" says "create": a word for a change.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Step 2 of onboarding a new app onto SparkPay. Applies answers.plan_catalog by minting real Stripe prices (same helper /api/apps/sync-plans uses) and registers/updates the app, then returns a downloadable integration kit (lib/spark-pay client + plans.json + sync script + INTEGRATION.md, modeled on still-applying's golden pattern) plus exact post_steps for env vars and secrets. Refuses to overwrite an existing app's plans unless answers.confirm_overwrite is true. Real secrets (a push-mode webhook secret) are only ever returned in post_steps, never written into a kit file.

    Annotations the server declares

    • title: Create App Kit
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: true
  • delete_couponDelete one coupon from an app.Reads text others writeNo findings

    Why it is in this group

    • The name "delete_coupon" says "delete": a word for removing or ending something.
    • The input takes a statement or command to run (input.code), which can change anything it can reach.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Delete one coupon from an app. WRITE, human-gated: a real delete requires an elicitation-capable client; the human sees the exact coupon (code, percent, redemption count) and must click approve. The model cannot confirm on its own, and confirm_used is only consulted for dry_run previews. Clients without elicitation are refused UNLESS the deployment sets SPARK_PAY_MCP_ALLOW_UNELICITED_WRITES=1 and the call passes confirm_write:true, which is how Claude Code and cron reach these writes; dry_run still works anywhere. Removes the Stripe coupon object and then the `coupons` row, exactly like the dashboard's delete, so no orphaned Stripe object is left behind. Deleting is permanent and the code becomes reusable afterwards. Set dry_run:true to see exactly which coupon would go (and its redemption count) without touching anything. Use list_coupons first to confirm the code.

    Annotations the server declares

    • title: Delete Coupon
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: true
    • openWorldHint: true
  • revoke_coupon_inviteWithdraw one address's right to spend a coupon: the targeted block.Reads text others writeNo findings

    Why it is in this group

    • The name "revoke_coupon_invite" says "revoke": a word for removing or ending something.
    • The input takes a statement or command to run (input.code), which can change anything it can reach.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Withdraw one address's right to spend a coupon: the targeted block. WRITE, human-gated: a real write requires an elicitation-capable client and the human sees the address and its current status; clients without elicitation are refused unless the deployment sets SPARK_PAY_MCP_ALLOW_UNELICITED_WRITES=1 and the call passes confirm_write:true (dry_run works anywhere). Works on an address that was never invited too, writing a revoked tombstone, so blocking someone does not depend on whether they were on the list and survives a later bulk re-invite. This only blocks FUTURE redemptions: if the status was already `redeemed` the discount has been granted, and undoing that means cancelling or refunding the subscription in the dashboard. Prefer this over delete_coupon when one recipient misbehaves -- deleting the code punishes every other invitee. Use list_coupon_invites first to see the current status.

    Annotations the server declares

    • title: Revoke Coupon Invite
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: true
    • openWorldHint: true
  • setup_pricing_pageGuided, stage-chained wizard that configures an app's hosted pricing page end to end: plan billing model + price (with clickable recommendations), plan name and badge (elite/recomm…Reads text others writeNo findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Guided, stage-chained wizard that configures an app's hosted pricing page end to end: plan billing model + price (with clickable recommendations), plan name and badge (elite/recommended), optional free tier, a seasonal coupon (SUMMER26-style code suggested from the current season), favicon-derived theme color + card style, and FAQ content. WRITE, human-gated per stage: each call runs AT MOST ONE elicitation form and commits only that stage, then returns next_call for the model to continue (stages: overview -> plans -> coupon -> branding -> content -> finish). The values the HUMAN types in each form OVERRIDE the arguments here - args only pre-fill form defaults (pass price_suggestions from market research and faq_drafts to seed them). Testimonials are NOT configurable here: they are collected from real customers on Sellular and rendered from there. Form stages are refused on clients without elicitation; overview/finish are read-only recon and work anywhere, as does dry_run (echoes the form + would-be writes, writes nothing). A declined stage leaves earlier committed stages applied - each stage is individually approved, and each effect is reversible via the standalone tools. Stripe prices are minted only by the plans stage with human-typed amounts (existing paid tiers only; free tiers never touch Stripe). After plan changes, run `bun run plans:sync` in consumer app repos. Use list_apps first for app_id.

    Annotations the server declares

    • title: Setup Pricing Page
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: true
  • update_plan_limitsChange the usage limits on an app's existing tiers (e.g.Reads text others writeNo findings

    Why it is in this group

    • The name "update_plan_limits" says "update": a word for a change.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Change the usage limits on an app's existing tiers (e.g. set ai_generations_daily to 200 on pro). WRITE, human-gated: a real write requires an elicitation-capable client; the human sees the per-tier before -> after diff and must click approve. Clients without elicitation are refused UNLESS the deployment sets SPARK_PAY_MCP_ALLOW_UNELICITED_WRITES=1 and the call passes confirm_write:true, which is how Claude Code and cron reach these writes; dry_run still works anywhere. Tier-keyed, so one call can set a new limit key across every tier in a single atomic write. Use -1 for unlimited, matching the catalog convention. This writes ONLY the limits object: it never touches Stripe, prices, or plan names, so it cannot mint a price by accident. Changing prices or adding a tier is a different job: use create_app_kit or the dashboard. Set dry_run to preview the before/after without writing. After this succeeds, run `bun run plans:sync` in the app repo and commit the regenerated plans.json so the code and the catalog agree.

    Annotations the server declares

    • title: Update Plan Limits
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: true
    • openWorldHint: true
  • update_plan_metaSet an existing tier's display metadata: its name and the two badge flags the public pricing page styles cards from: is_elite (gold/carbon BEST VALUE card with the rainbow title an…Reads text others writeNo findings

    Why it is in this group

    • The name "update_plan_meta" says "update": a word for a change.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Set an existing tier's display metadata: its name and the two badge flags the public pricing page styles cards from: is_elite (gold/carbon BEST VALUE card with the rainbow title and app icon) and recommended (MOST POPULAR). A tier called "Elite" is NOT styled as elite unless is_elite is actually true on the stored plan; the page reads the flag, never the tier slug, so a catalog whose top tier lacks the flag renders as a plain grey card. Use this to repair that. WRITE, human-gated: the human sees the before -> after diff and approves (clients without elicitation pass through, since this touches no money: it never reaches Stripe, never changes an amount, limits or billing model, and is reversible by calling it again). Setting a flag false REMOVES the key rather than storing false, keeping exported plans.json minimal. Convention is at most ONE is_elite and ONE recommended tier per app: this tool does not clear the flag from its siblings, so unset the old holder in a second call. Use list_plans first for tier slugs and current flags, and run `bun run plans:sync` in the app repo afterwards to regenerate plans.json.

    Annotations the server declares

    • title: Update Plan Meta
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: true
    • openWorldHint: true
  • update_plan_pricingChange what an app's EXISTING tiers charge: the amount, the billing model (one_time vs subscription), or both.Reads text others writeNo findings

    Why it is in this group

    • The name "update_plan_pricing" says "update": a word for a change.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Change what an app's EXISTING tiers charge: the amount, the billing model (one_time vs subscription), or both. WRITE, human-gated at the highest tier: a real write requires an elicitation-capable client, and the human types the tier, billing model and price into a form. Those typed values OVERRIDE the arguments here, which only pre-fill the proposal, so the model cannot move a price on its own. Clients without elicitation are refused UNLESS the deployment sets SPARK_PAY_MCP_ALLOW_UNELICITED_WRITES=1 and the call passes confirm_write:true, which is how Claude Code and cron reach these writes; dry_run still works anywhere. Stripe prices are immutable, so this mints a NEW price and repoints the plan; the old price object is left intact and existing subscribers keep the price they signed up on, so this changes what NEW buyers pay. Switching a tier to one_time also clears its monthly/yearly price ids, so checkout cannot fall back to a recurring price the pricing page no longer advertises. Free, contact and payg tiers are refused: converting those is a catalog change, not a reprice. Use list_plans first for tier slugs and current amounts, and run `bun run plans:sync` in the app repo afterwards to regenerate plans.json.

    Annotations the server declares

    • title: Update Plan Pricing
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: true
Can't be undone 1Claude Code: ask
  • invite_to_couponAdd email addresses to a coupon's invite list and, by default, switch the coupon to invite-only so the list is actually enforced.Reads text others writeCan send data outNo findings

    Why it is in this group

    • The name "invite_to_coupon" says "invite": a word for sending, and what is sent cannot be taken back.
    • The input takes a statement or command to run (input.code), which can change anything it can reach.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    • outboundThe name "invite_to_coupon" says "invite": it sends or publishes something to other people.
    Description, as the server wrote it
    Add email addresses to a coupon's invite list and, by default, switch the coupon to invite-only so the list is actually enforced. WRITE, human-gated: a real write requires an elicitation-capable client and the human sees every address plus whether the coupon is about to be closed to everyone else; clients without elicitation are refused unless the deployment sets SPARK_PAY_MCP_ALLOW_UNELICITED_WRITES=1 and the call passes confirm_write:true (dry_run works anywhere). Addresses are lowercased and de-duplicated, malformed ones are reported rather than dropped, and an address already on the list is left exactly as it is -- so re-inviting someone who already redeemed does NOT hand them a second free account, and re-inviting someone who was revoked does NOT un-block them. Set enforce_invite_only:false only to stage a list before gating the code; the result then carries a warning that the coupon is still open. Use dry_run:true first to see exactly who would be added.

    Annotations the server declares

    • title: Invite To Coupon
    • readOnlyHint: false
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
Changes things 7Claude Code: ask
  • bulk_create_couponCreate the SAME discount code across many apps in one call: the portfolio-promo shortcut (e.g.Reads text others writeNo findings

    Why it is in this group

    • The name "bulk_create_coupon" says "create": a word for a change.
    • The input takes a statement or command to run (input.code), which can change anything it can reach.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Create the SAME discount code across many apps in one call: the portfolio-promo shortcut (e.g. a summer sale on every owned app). WRITE, human-gated: a real create requires an elicitation-capable client; the human enters the coupon terms ONCE in a form (overriding the arguments here) and that one approval covers the whole batch. Clients without elicitation are refused UNLESS the deployment sets SPARK_PAY_MCP_ALLOW_UNELICITED_WRITES=1 and the call passes confirm_write:true, which is how Claude Code and cron reach these writes; dry_run still works anywhere. Runs each app through the same fully-guarded create as create_coupon. Unlike a single call it does NOT fail the batch when an app already has the code: that app is reported `skipped` and the rest are still created. Returns a per-app outcome matrix plus created/skipped/failed counts. Use dry_run:true first to see which apps would create vs skip. Pass explicit app_ids (from list_apps). There is deliberately no 'all apps' shortcut, so a discount can't hit an app you didn't name.

    Annotations the server declares

    • title: Bulk Create Coupon
    • readOnlyHint: false
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • create_couponCreate a percentage discount coupon for an app.Reads text others writeNo findings

    Why it is in this group

    • The name "create_coupon" says "create": a word for a change.
    • The input takes a statement or command to run (input.code), which can change anything it can reach.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Create a percentage discount coupon for an app. WRITE, human-gated: a real create requires an elicitation-capable client (e.g. Claude Desktop). The human enters the final code, percent, expiry, visibility, and tier scope in a form; those values OVERRIDE the arguments here, which only pre-fill the form message as a proposal. Clients without elicitation are refused UNLESS the deployment sets SPARK_PAY_MCP_ALLOW_UNELICITED_WRITES=1 and the call passes confirm_write:true, which is how Claude Code and cron reach these writes; dry_run still works anywhere. Inserts a `coupons` row and best-effort-mints the matching Stripe coupon (checkout backfills it on first use if Stripe is unreachable). The code is uppercased and must be unique per app. By default the coupon applies to every paid tier. Set dry_run:true first to preview the exact row without writing. Cannot discount a free-only app. Three house rules are enforced and cannot be argued out of: every coupon expires within 6 months (omitting expires_at means the 6-month term, not a code that never expires); a coupon at 90%+ is created invite_only, so it is redeemable only by addresses added with invite_to_coupon; and a 90%+ coupon is pinned to the second-highest paid recurring tier and may never cover the top tier or a one-time (lifetime) plan. Use list_coupons afterwards to confirm, and list_plans to see an app's tier slugs.

    Annotations the server declares

    • title: Create Coupon
    • readOnlyHint: false
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • list_coupon_invitesList who a coupon was offered to and what they did with it: per-address status (invited / redeemed / revoked), where the invite came from, and the timestamps.Reads text others write1 finding

    Why it is in this group

    • The name "list_coupon_invites" says "list": a word for reading.
    • The input takes a statement or command to run (input.code), which can change anything it can reach.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.

    Findings

    • Warn

      The server marks "list_coupon_invites" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

      Matched text

      annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":true}; The name "list_coupon_invites" says "list": a word for reading.

      What this check looks for
    Description, as the server wrote it
    List who a coupon was offered to and what they did with it: per-address status (invited / redeemed / revoked), where the invite came from, and the timestamps. Also returns the coupon's own `invite_only` flag -- read that FIRST: when it is false the invite rows are recorded but NOT enforced, and anyone holding the code can still redeem it. Read-only. Use list_coupons for the coupon's terms, invite_to_coupon to add addresses, revoke_coupon_invite to withdraw one.

    Annotations the server declares

    • title: List Coupon Invites
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • plan_app_kitStep 1 of onboarding a new app onto SparkPay.Reads text others writeNo findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Step 1 of onboarding a new app onto SparkPay. Recons app_name against existing apps and the rest of the portfolio's plan catalogs, then returns automated findings plus the minimum set of questions that genuinely need a human answer (plan prices, billing model, redirect URLs, trial length, push-webhook opt-in). Pass the answers to create_app_kit next.

    Annotations the server declares

    • title: Plan App Kit
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • revenue_by_appPer-app rollup in USD cents (latest_payments/refunded/MRR + paid customer count), largest refund-adjusted latest payments first.Reads text others writeNo findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Per-app rollup in USD cents (latest_payments/refunded/MRR + paid customer count), largest refund-adjusted latest payments first. latest_payments sums the most recent payment per customer, so it is a run-rate and NOT lifetime turnover. No lifetime figure is available per app (accrual is recorded per owner). Use to compare products at a glance; use revenue_summary for one app in depth.

    Annotations the server declares

    • title: Revenue By App
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • revenue_summaryRevenue and subscriber summary in USD cents (normalized from each transaction currency).Reads text others writeNo findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Revenue and subscriber summary in USD cents (normalized from each transaction currency). Returns latest_payments (the MOST RECENT payment summed per customer: a run-rate, NOT lifetime turnover, because amount_paid is overwritten on renewal), refunds, approximate MRR, active/trialing/paid counts, and lifetime_net (the only cumulative total held, accrued per OWNER across all their apps and net of refunds; null when no accrual record exists). lifetime_net counts only charges whose Stripe webhook reached this app, so a missing renewal event under-counts it silently. A 0 or a low figure means "not recorded here", never "no sales". For turnover, accounts or tax, read Stripe directly: it is authoritative, and this database cannot reconstruct lifetime gross. Optionally scope by app_id; omit for platform-wide totals.

    Annotations the server declares

    • title: Revenue Summary
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • set_coupon_listedShow, hide or SCHEDULE an EXISTING coupon on the public pricing pages by setting its `listed` flag and optional `listed_from` start date.Reads text others writeNo findings

    Why it is in this group

    • The name "set_coupon_listed" says "set": a word for a change.
    • The input takes a statement or command to run (input.code), which can change anything it can reach.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Show, hide or SCHEDULE an EXISTING coupon on the public pricing pages by setting its `listed` flag and optional `listed_from` start date. WRITE, human-gated: a real write requires an elicitation-capable client and the human sees the code, the discount, which way it is moving and any start date; clients without elicitation pass through, since this only moves a display flag (dry_run works anywhere). These are deliberately the ONLY fields an agent can change on a live coupon: the percent, expiry and tier scope are terms somebody may already have been quoted, while `listed`/`listed_from` are purely a display decision. Together they decide whether the public coupons API returns the code today, which is what every app's promo bar draws from; the discount itself does not move and no Stripe object is touched. `listed_from` in the future STAGES a seasonal coupon so it appears on its own date with no cron: it never blocks anyone holding the code from redeeming it, it only holds the code back from the promo bar. Listing with no `listed_from` clears any existing schedule (lists now). Listing an INVITE-ONLY coupon is refused outright, since that would advertise a near-free code to people who cannot redeem it. A `listed_from` with listed:false is refused as a contradiction. Listing an expired/maxed coupon, or scheduling a start at or after expiry, is allowed but warns. A coupon already in the requested state reports `noop`. Use list_coupons first to see codes and their current visibility.

    Annotations the server declares

    • title: Set Coupon Listed
    • readOnlyHint: false
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
Only reads 8Show 8 read-only tools
  • get_customerLook up a customer by email.Reads private dataReads text others writeNo findings

    Why it is in this group

    • The name "get_customer" says "get": a word for reading.

    What it can reach

    • private-dataThe description's first sentence says "email": it reads mail.
    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    • untrusted-inputThe description's first sentence says "email": it reads inbound email, which other people can write.
    Description, as the server wrote it
    Look up a customer by email. Returns every subscription/registration for that email across apps (or within one app if app_id is given), with full detail: verification, plan, billing period, usage, refunds. Empty when the email is unknown.

    Annotations the server declares

    • title: Get Customer
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • list_appsList the registered apps (app_id + name).Reads text others writeNo findings

    Why it is in this group

    • The name "list_apps" says "list": a word for reading.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    List the registered apps (app_id + name). Use this first to discover valid app_id values, then pass one to the other tools to scope results to a single product.

    Annotations the server declares

    • title: List Apps
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • list_couponsList discount coupons (code, percent off, expiry, max uses, use count, which paid tiers they apply to).Reads text others writeNo findings

    Why it is in this group

    • The name "list_coupons" says "list": a word for reading.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    List discount coupons (code, percent off, expiry, max uses, use count, which paid tiers they apply to). Optionally scope by app_id and/or restrict to currently-active ones (not expired, not maxed out). Read-only: cannot create, edit, or delete coupons.

    Annotations the server declares

    • title: List Coupons
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • list_plansDownload an app's live plan catalog: tiers, names, prices, trial, and the machine-readable `limits` that gate features inside the app.Reads text others writeNo findings

    Why it is in this group

    • The name "list_plans" says "list": a word for reading.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Download an app's live plan catalog: tiers, names, prices, trial, and the machine-readable `limits` that gate features inside the app. This is the source of truth each app repo's `bun run plans:sync` regenerates lib/spark-pay/plans.json from, so diff against this rather than trusting the committed file. Stripe price/product IDs are omitted unless include_stripe_ids is set.

    Annotations the server declares

    • title: List Plans
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • list_recent_cancellationsRecently canceled paid subscriptions (excludes free rows), most recent cancellation first, with what each customer was paying.Reads text others writeNo findings

    Why it is in this group

    • The name "list_recent_cancellations" says "list": a word for reading.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Recently canceled paid subscriptions (excludes free rows), most recent cancellation first, with what each customer was paying. Use for churn review; list_subscriptions orders by signup date instead.

    Annotations the server declares

    • title: List Recent Cancellations
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • list_recent_paymentsList recent paid transactions (excludes free-tier rows), newest first, ordered by payment date.Reads text others writeNo findings

    Why it is in this group

    • The name "list_recent_payments" says "list": a word for reading.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    List recent paid transactions (excludes free-tier rows), newest first, ordered by payment date. Optionally scope by app_id. Use to review recent sales activity. Amounts are in the transaction currency (cents).

    Annotations the server declares

    • title: List Recent Payments
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • list_referralsList referral records (who referred whom, reward coupon code, pending/rewarded status) with what the referred buyer paid, newest first.Reads text others writeNo findings

    Why it is in this group

    • The name "list_referrals" says "list": a word for reading.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    List referral records (who referred whom, reward coupon code, pending/rewarded status) with what the referred buyer paid, newest first. Use for working out creator commission: a one_time referredPaymentType means referredAmountPaidCents is the whole sale; for a recurring row it is the LATEST payment, not lifetime, so do not commission on it. Optionally scope by app_id and/or status.

    Annotations the server declares

    • title: List Referrals
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true
  • list_subscriptionsList subscriptions/registrations, newest first.Reads text others writeNo findings

    Why it is in this group

    • The name "list_subscriptions" says "list": a word for reading.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    List subscriptions/registrations, newest first. Optionally filter by app_id, status, or payment_type. Returns metadata (email, status, plan, amounts). Use get_customer for a single user, or revenue_summary for totals.

    Annotations the server declares

    • title: List Subscriptions
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: true

This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.

Exfil surface

Which tools read private data, which read text someone else can write, and which can send data out. One of each is a path a prompt injection can use.

25 paths in this server alone from private data, through text someone else wrote, to a way out. Denying 1 outbound tool breaks every one.

Reads private data1
Reads files, mail, messages, databases, secrets, calendars or repositories.
Reads text others write25
Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
Can send data out1
Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL.
deny to break every path
{
  "permissions": {
    "deny": [
      "mcp__sparkpay__invite_to_coupon"
    ]
  }
}

Other servers in the same agent add their own tools to these paths. Audit your whole setup to see them together.

Remote posture

What anyone on the internet sees at the remote endpoint: an anonymous initialize and tools/list, and the public OAuth metadata.

Tools listed without sign-in

Looked at Nov 20, 2026, 18:52 UTC, unchanged since Nov 20, 2026. No credentials were sent and no tool was called. What is looked at.

Reachable
yes
HTTP status of an anonymous initialize
200
tools/list served without credentials
yes
Tools that change things, listed without credentials
19
Protocol version
2025-06-18
WWW-Authenticate on 401
not observed
resource_metadata in WWW-Authenticate
not observed
RFC 9728 protected resource metadata
not observed
Authorization servers
not observed
Legacy HTTP+SSE transport
no
Dynamic client registration
not observed
Client ID metadata documents
not observed
  • HighThe whole server

    The endpoint served tools/list without credentials, and the list has 19 tools that change things (9 destructive, 1 irreversible, 9 write). Whether a call also works without credentials was not tested: tools are never called.

    Matched text

    https://sparkpay.dev/api/mcp/mcp: tools/list returned 27 tools to an anonymous client, including delete_coupon (destructive), bulk_delete_coupon (destructive), revoke_coupon_invite (destructive), create_app_kit (destructive), update_plan_limits (destructive) and 14 more

    What this check looks for

Posture changes

  1. Nov 20, 2026
    • Tools that change things, listed without credentials: 17 to 19

Findings

Matches of fixed checks, with the text they matched.

1 warn
  • The server marks "list_coupon_invites" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

    Matched text

    annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":true}; The name "list_coupon_invites" says "list": a word for reading.

    What this check looks for

History

Changes to its tool definitions and to its remote endpoint, newest first.

  1. PostureRemote posture changed. Tools that change things, listed without credentials: 17 -> 19.View diff
    • Tools that change things, listed without credentials: 17 to 19
  2. Tools10 changed.~10 changedView diff
  3. Tools25 changed.~25 changedView diff

Install

Pinned to version 1.0.0. Values in angle brackets are yours to fill in.

claude mcp add --transport http sparkpay https://sparkpay.dev/api/mcp/mcp
claude mcp add --transport http sparkpay https://sparkpay.dev/api/mcp/mcp

Run in a terminal.

Claude Code permissions

Allow the reads, ask before a change or a way out, deny the deletes. Paste into .claude/settings.json.

.claude/settings.json
{
  "permissions": {
    "allow": [
      "mcp__sparkpay__list_apps",
      "mcp__sparkpay__list_subscriptions",
      "mcp__sparkpay__get_customer",
      "mcp__sparkpay__list_recent_payments",
      "mcp__sparkpay__list_referrals",
      "mcp__sparkpay__list_recent_cancellations",
      "mcp__sparkpay__list_coupons",
      "mcp__sparkpay__list_plans"
    ],
    "ask": [
      "mcp__sparkpay__revenue_summary",
      "mcp__sparkpay__revenue_by_app",
      "mcp__sparkpay__create_coupon",
      "mcp__sparkpay__bulk_create_coupon",
      "mcp__sparkpay__set_coupon_listed",
      "mcp__sparkpay__list_coupon_invites",
      "mcp__sparkpay__invite_to_coupon",
      "mcp__sparkpay__plan_app_kit"
    ],
    "deny": [
      "mcp__sparkpay__delete_coupon",
      "mcp__sparkpay__bulk_delete_coupon",
      "mcp__sparkpay__revoke_coupon_invite",
      "mcp__sparkpay__create_app_kit",
      "mcp__sparkpay__update_plan_limits",
      "mcp__sparkpay__update_plan_meta",
      "mcp__sparkpay__update_plan_pricing",
      "mcp__sparkpay__archive_plan",
      "mcp__sparkpay__setup_pricing_page"
    ]
  }
}

Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. A read that can send data out (it takes a URL, say) asks first too. It assumes the server is installed as sparkpay. Review it before you use it.

Capability tags behind 25 rules
  • mcp__sparkpay__list_apps untrusted-input
  • mcp__sparkpay__list_subscriptions untrusted-input
  • mcp__sparkpay__get_customer private-data, untrusted-input
  • mcp__sparkpay__list_recent_payments untrusted-input
  • mcp__sparkpay__revenue_summary untrusted-input
  • mcp__sparkpay__revenue_by_app untrusted-input
  • mcp__sparkpay__list_referrals untrusted-input
  • mcp__sparkpay__list_recent_cancellations untrusted-input
  • mcp__sparkpay__list_coupons untrusted-input
  • mcp__sparkpay__create_coupon untrusted-input
  • mcp__sparkpay__bulk_create_coupon untrusted-input
  • mcp__sparkpay__delete_coupon untrusted-input
  • mcp__sparkpay__bulk_delete_coupon untrusted-input
  • mcp__sparkpay__set_coupon_listed untrusted-input
  • mcp__sparkpay__list_coupon_invites untrusted-input
  • mcp__sparkpay__invite_to_coupon untrusted-input, outbound
  • mcp__sparkpay__revoke_coupon_invite untrusted-input
  • mcp__sparkpay__plan_app_kit untrusted-input
  • mcp__sparkpay__create_app_kit untrusted-input
  • mcp__sparkpay__list_plans untrusted-input
  • mcp__sparkpay__update_plan_limits untrusted-input
  • mcp__sparkpay__update_plan_meta untrusted-input
  • mcp__sparkpay__update_plan_pricing untrusted-input
  • mcp__sparkpay__archive_plan untrusted-input
  • mcp__sparkpay__setup_pricing_page untrusted-input

Pin

Keep the decision: fail CI when the tools change, follow every change, or show a badge.

Pin in CI

Fail a build when the tools you reviewed are not the tools the server serves today.

curl -s "https://radar.anlyon.com/api/v1/servers/dev.sparkpay/sparkpay/verify?hash=8b924639839c15c332bced2f4a6b181e19cd96011afe93f305e5734458af013a" | jq -e '.changed == false'

Pinned hash 8b924639839c, from Nov 15, 2026, 03:18 UTC.

Watch

Get told when its tools change: a new tool, a new description, a new class.

https://radar.anlyon.com/servers/dev.sparkpay/sparkpay/changes.xml

An Atom feed. Works in any feed reader or Slack's RSS app.

Badge

Anlyon Radar badge for dev.sparkpay/sparkpay
[![Anlyon Radar](https://radar.anlyon.com/badge/dev.sparkpay/sparkpay.svg)](https://radar.anlyon.com/servers/dev.sparkpay/sparkpay)

Source: MCP Registry. Synced Oct 21, 2026, 18:12 UTC.

SparkPay MCP server: questions and answers

Answered from the scan above, not from the publisher's description.

Is the SparkPay MCP server safe?

No scan can promise that. What its definitions show: 25 tools, 17 of which can change, send or delete something, and 1 finding from fixed checks. Allow the reads, ask before each change, and deny the deletes; the permissions block on this page does that.

Can the SparkPay MCP server delete data?

Yes. 9 tools can delete, cancel or revoke something: delete_coupon, bulk_delete_coupon, revoke_coupon_invite, create_app_kit, update_plan_limits, update_plan_meta and 3 more.

Can the SparkPay MCP server do things that cannot be undone?

Yes. 1 tool can send, pay, publish or merge: invite_to_coupon.

What tools does the SparkPay MCP server have?

25 tools: 9 destructive, 1 irreversible, 7 write, 8 read. delete_coupon, bulk_delete_coupon, revoke_coupon_invite, create_app_kit, update_plan_limits, update_plan_meta, update_plan_pricing, archive_plan and 17 more.

Does the SparkPay MCP server need sign-in or OAuth?

Its remote endpoint listed its tools to an anonymous client, without sign-in.

Does the SparkPay MCP server send data to outside services?

1 tool can carry data out (it sends, posts or publishes, or takes a recipient, channel or URL): invite_to_coupon.

Has the SparkPay MCP server changed its tools?

Yes, 2 times since Anlyon Radar started tracking it, most recently on Nov 15, 2026. Every change is on this page and in its Atom feed.