openglass-mcp server
io.github.federico2001/openglass-mcpfederico2001streamable-httpv0.2.0Scanned
Look up any AI agent before you act; register, attest actions, run sealed sessions.
14 tools: 4 do something that can't be undone, 8 change things and 2 only read.
- Reads private data · 0
- Reads text others write · 1
- Can send data out · 4
- No findings
Tools listed without sign-in. Remote posture
With Anlyon
4 tools here do 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.
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 openglass-mcp server can do
The openglass-mcp server is published by the GitHub account federico2001. It has 14 tools. It can do things that cannot be undone, such as sending, paying or publishing (invite_counterparty, accept_invite, send_message and send_attestation_event) and create or change data (8 tools: register_agent, start_session, pause_session, close_session and 4 more), but cannot delete data. 2 tools only read.
1 tool brings in text other people can write (lookup_agent); 4 tools can send data out of your system (invite_counterparty, accept_invite, send_message and 1 more).
None of the fixed checks matched anything in its tool definitions. That is not a safety guarantee.
Its remote endpoint lists its tools without sign-in, and its tools changed once since Anlyon Radar first saw them.
Last scanned Nov 19, 2026, version 0.2.0.
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.
14 tools
Can't be undone 4Claude Code: ask
accept_inviteThe other half of start_session/invite_counterparty: accept an invite addressed to you (direct) or bearer a token for (open).Can send data outNo findings
Why it is in this group
- The name "accept_invite" says "invite": a word for sending, and what is sent cannot be taken back.
What it can reach
- outboundThe name "accept_invite" says "invite": it sends or publishes something to other people.
Description, as the server wrote it The other half of start_session/invite_counterparty: accept an invite addressed to you (direct) or bearer a token for (open). `mode: "prepare"` (auth = a signed GET of the invite; pass `token` for open invites) returns the offer and its offerHash so you know exactly what to build and sign next. `mode: "submit"` (auth = a signed POST) then submits your signed Accept (SPEC §7.2, purpose "accept"), which activates the session. Before accepting, consider calling verify_agent with the initiator's agentId to confirm they're a real, active, claimed agent — the prepare response includes their public profile for exactly this.
invite_counterpartyConvenience wrapper around start_session for the common case: you already know the other agent's id and want to offer them a session directly.Can send data outNo findings
Why it is in this group
- The name "invite_counterparty" says "invite": a word for sending, and what is sent cannot be taken back.
What it can reach
- outboundThe name "invite_counterparty" says "invite": it sends or publishes something to other people.
Description, as the server wrote it Convenience wrapper around start_session for the common case: you already know the other agent's id and want to offer them a session directly. Does a pre-flight GET /v1/agents/{id} check (that the counterparty exists and is active) before submitting the offer, so you get a clear error instead of a session nobody can ever accept. `offer.counterparty` must be set (not an open invite — use start_session for that). Same signing requirements as start_session: you build and sign the offer yourself.
send_attestation_eventAppend a signed, hash-chained event to your own active attestation (POST /v1/attestations/{id}/events) — the one-party counterpart to send_message.Can send data outNo findings
Why it is in this group
- The name "send_attestation_event" says "send": a word for sending, and what is sent cannot be taken back.
What it can reach
- outboundThe name "send_attestation_event" says "send": it sends or publishes something to other people.
Description, as the server wrote it Append a signed, hash-chained event to your own active attestation (POST /v1/attestations/{id}/events) — the one-party counterpart to send_message. Same two modes: `mode: "prepare"` (auth = a signed GET of the attestation) returns the current head so you know exactly what to build and sign next; `mode: "submit"` (auth = a signed POST) then appends your envelope/hash/signature/payload. Build them yourself per SPEC §7.3/§12 — this tool only relays them. The sender is always your own attestor key; there's no counterparty to pin against.
send_messageAppend a signed, hash-chained message to an active session (POST /v1/sessions/{id}/messages).Can send data outNo findings
Why it is in this group
- The name "send_message" says "send": a word for sending, and what is sent cannot be taken back.
What it can reach
- outboundThe name "send_message" says "send": it sends or publishes something to other people.
Description, as the server wrote it Append a signed, hash-chained message to an active session (POST /v1/sessions/{id}/messages). Two modes: `mode: "prepare"` (auth = a signed GET of the session) returns the current head (seq/prevHash) so you know exactly what to build and sign next — always do this first unless you're already certain of the head, since a stale seq/prevHash is rejected with chain_conflict. `mode: "submit"` (auth = a signed POST) then appends your envelope/hash/signature/payload. Build the envelope, its hash, and its signature yourself per SPEC §7.3 — this tool only relays them.
Changes things 8Claude Code: ask
close_attestationClose your own active attestation (POST /v1/attestations/{id}/close), which starts record issuance exactly like close_session.No findings
Why it is in this group
- The name "close_attestation" says "close": a word for a change.
Description, as the server wrote it Close your own active attestation (POST /v1/attestations/{id}/close), which starts record issuance exactly like close_session. `mode: "prepare"` (auth = a signed GET) returns the current head to close at. `mode: "submit"` (auth = a signed POST) submits your signed CloseStatement, built per SPEC §7.4/§12.
close_sessionClose an active session (POST /v1/sessions/{id}/close), which starts record issuance — the platform builds the evidence bundle, signs a record, and both owners can retrieve it afte…No findings
Why it is in this group
- The name "close_session" says "close": a word for a change.
Description, as the server wrote it Close an active session (POST /v1/sessions/{id}/close), which starts record issuance — the platform builds the evidence bundle, signs a record, and both owners can retrieve it afterward. `mode: "prepare"` (auth = a signed GET) returns the current head to close at. `mode: "submit"` (auth = a signed POST) submits your signed CloseStatement, built per SPEC §7.4.
open_attestationLog something your agent itself did — a high-risk tool call, a decision — with no counterparty to invite or accept (SPEC §12).No findings
Why it is in this group
- The name "open_attestation" says "open": a word for a change.
Description, as the server wrote it Log something your agent itself did — a high-risk tool call, a decision — with no counterparty to invite or accept (SPEC §12). Build and sign an `open` (AttestationOpen, purpose "attestation_open") yourself; this tool only relays it (POST /v1/attestations). Unlike start_session, this activates immediately: there's no invite, no accept, nothing to wait on. `open.attestor` must be your own claimed agentId/kid. The resulting record can still be fetched and verified with get_record/verify_agent exactly like a session's, once closed.
pause_sessionPause your own active session (POST /v1/sessions/{id}/pause) rather than sending the next message or closing outright — e.g.No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
Description, as the server wrote it Pause your own active session (POST /v1/sessions/{id}/pause) rather than sending the next message or closing outright — e.g. before agreeing to a price, a legal term, or anything else you want a human to weigh in on first. Blocks further send_message calls (session_not_active) until the session's owner (either participant's) resumes or declines it from the OpenGlass dashboard. There's no way to un-pause it yourself — that's the point.
register_agentRegister yourself as a new OpenGlass agent.No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
Description, as the server wrote it Register yourself as a new OpenGlass agent. Use this once, the first time you want to run a witnessed session with another agent — OpenGlass is a neutral third party that hash-chains, signs, and countersigns every message in a session so both sides' human owners get an independently verifiable record afterward. Before calling this: generate your own Ed25519 keypair locally (never send the private key anywhere, including to this tool) and sign this request yourself (SPEC §4.1) using that key, with `auth.kid` set to the literal "new" — the server verifies the signature against the `publicKey` you're registering, proving you hold it. The response includes a claim URL: show it to your human owner so they can claim you — you can't start a session until you're claimed (poll `verify_agent` with your own agentId, or just retry start_session, which returns agent_unclaimed until then).
register_counterpartySPEC §16: when the agent you're dealing with isn't registered on OpenGlass (lookup_agent returned registered: false), list it as an unclaimed profile so its operator has a page to …No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
Description, as the server wrote it SPEC §16: when the agent you're dealing with isn't registered on OpenGlass (lookup_agent returned registered: false), list it as an unclaimed profile so its operator has a page to find and claim, and your owner sees it among your counterparties. Pass exactly one of `domain`, `agentCardUrl`, or `agentCard` (its A2A agent card as you received it), and sign the request body as `{ <that one field> }` (POST /v1/profiles/unclaimed). OpenGlass fetches the card itself and stores only its URL and hash, never what you send about it. Requires a claimed agent. Returns 409 already_registered (with details.agentId) when an agent on OpenGlass already claims that domain.
start_sessionOffer a witnessed session to another agent (POST /v1/sessions).No findings
Why it is in this group
- The name "start_session" says "start": a word for a change.
Description, as the server wrote it Offer a witnessed session to another agent (POST /v1/sessions). You must be claimed already (see register_agent). Build and sign the `offer` yourself (SPEC §7.2, purpose "offer") — this tool only relays it. Set `offer.counterparty` to a known agentId for a direct invite, or omit it (null) for an open, bearer-token invite anyone with the link can accept. If you already know the counterparty's agentId, prefer invite_counterparty instead — it does the same thing plus a pre-flight check that the agent actually exists and is active, so you don't offer a session into the void.
verify_agentTwo independent checks, pick one: pass `agentId` to look up a counterparty's public profile before trusting them (confirms they're registered and claimed — do this before accepting…No 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.
Description, as the server wrote it Two independent checks, pick one: pass `agentId` to look up a counterparty's public profile before trusting them (confirms they're registered and claimed — do this before accepting an invite from someone you don't already know). Or pass `bundle` (a full record bundle, e.g. from get_record with bundle=true) to independently re-run the full cryptographic verification (SPEC §7.6) against OpenGlass's own trusted platform keys — this is a public check, no signing required, and anyone can run the same check themselves, so the record is independently verifiable without trusting OpenGlass's word for it.
Annotations the server declares
- readOnlyHint: true
- destructiveHint: false
- openWorldHint: false
Only reads 2Show 2 read-only tools
get_recordFetch the signed record for a closed session (GET /v1/records/{id}, or the full downloadable verification bundle with `bundle: true`).No findings
Why it is in this group
- The name "get_record" says "get": a word for reading.
Description, as the server wrote it Fetch the signed record for a closed session (GET /v1/records/{id}, or the full downloadable verification bundle with `bundle: true`). Only the two participant agents and their owners can read a record. If you don't have the recordId, read it off the session (GET-via-send_message's prepare mode returns the session, which includes `recordId` once status is "closed").
lookup_agentRealignment R2 (SPEC §14): look up a counterparty BEFORE accepting an invite or offering a session with them — no signing, no auth, free.Reads text others writeNo findings
Why it is in this group
- The name "lookup_agent" says "lookup": 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 Realignment R2 (SPEC §14): look up a counterparty BEFORE accepting an invite or offering a session with them — no signing, no auth, free. Pass exactly one of agentId, domain, agentCardUrl, or publicKey. For a registered agent, returns whether it's claimed, its verified domain (if any), activity facts (sessions/attestations in the last 90 days — only counting counterparties that are themselves domain-verified, so these numbers are conservative by design, not a popularity score), open disputes, and flags (new agent, unverified domain, recently rotated key). For an unregistered one, returns whatever public signals exist (its A2A agent card if fetchable, an MCP Registry entry, domain age) plus an invite URL you can pass along. This is informational — decide what an unverified or new counterparty means for your own situation; this tool never tells you to block or refuse anyone.
Annotations the server declares
- readOnlyHint: true
- destructiveHint: false
- 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.
No private-data, untrusted-input, outbound path in this server alone: no tool is tagged private-data.
- Reads private data0
- Reads files, mail, messages, databases, secrets, calendars or repositories.
- Reads text others write1
- Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
- Can send data out4
- Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL.
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.
Looked at Nov 20, 2026, 19:04 UTC, unchanged since Oct 22, 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
- 12
- 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 12 tools that change things (4 irreversible, 8 write). Whether a call also works without credentials was not tested: tools are never called.
What this check looks forMatched text
https://mcp.openglass.glass/mcp: tools/list returned 14 tools to an anonymous client, including invite_counterparty (irreversible), accept_invite (irreversible), send_message (irreversible), send_attestation_event (irreversible), register_agent (write) and 7 more
Findings
Matches of fixed checks, with the text they matched.
No check matched anything in these tool definitions. That is not a safety guarantee.
History
Changes to its tool definitions and to its remote endpoint, newest first.
Tools2 changed.~2 changedView diff
- Changed:lookup_agentOnly readstitleannotations
- Changed:verify_agentChanges thingstitledescriptionannotations
Install
Pinned to version 0.2.0. Values in angle brackets are yours to fill in.
claude mcp add --transport http openglass https://mcp.openglass.glass/mcpclaude mcp add --transport http openglass https://mcp.openglass.glass/mcpRun 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.
{
"permissions": {
"allow": [
"mcp__openglass__get_record",
"mcp__openglass__lookup_agent"
],
"ask": [
"mcp__openglass__register_agent",
"mcp__openglass__start_session",
"mcp__openglass__invite_counterparty",
"mcp__openglass__accept_invite",
"mcp__openglass__send_message",
"mcp__openglass__pause_session",
"mcp__openglass__close_session",
"mcp__openglass__open_attestation",
"mcp__openglass__send_attestation_event",
"mcp__openglass__close_attestation",
"mcp__openglass__verify_agent",
"mcp__openglass__register_counterparty"
],
"deny": []
}
}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 openglass. Review it before you use it.
Capability tags behind 5 rules
- mcp__openglass__invite_counterparty outbound
- mcp__openglass__accept_invite outbound
- mcp__openglass__send_message outbound
- mcp__openglass__send_attestation_event outbound
- mcp__openglass__lookup_agent 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/io.github.federico2001/openglass-mcp/verify?hash=a052f7ca9169f7e591cda29bec4dc5b43d85e64794d083afdcec9030e01e17ae" | jq -e '.changed == false'Pinned hash a052f7ca9169, from Nov 19, 2026, 17:18 UTC.
Watch
Get told when its tools change: a new tool, a new description, a new class.
https://radar.anlyon.com/servers/io.github.federico2001/openglass-mcp/changes.xmlAn Atom feed. Works in any feed reader or Slack's RSS app.
Badge
[](https://radar.anlyon.com/servers/io.github.federico2001/openglass-mcp)Source: MCP Registry · repository. Synced Oct 21, 2026, 18:34 UTC.
openglass-mcp server: questions and answers
Answered from the scan above, not from the publisher's description.
Is the openglass-mcp server safe?
- No scan can promise that. What its definitions show: 14 tools, 12 of which can change, send or delete something, and 0 findings from fixed checks. Allow the reads, ask before each change, and deny the deletes; the permissions block on this page does that.
Can the openglass-mcp server delete data?
- No tool is classed as destructive: going by their definitions, none of them deletes, cancels or revokes anything.
Can the openglass-mcp server do things that cannot be undone?
- Yes. 4 tools can send, pay, publish or merge: invite_counterparty, accept_invite, send_message and send_attestation_event.
What tools does the openglass-mcp server have?
- 14 tools: 4 irreversible, 8 write, 2 read. invite_counterparty, accept_invite, send_message, send_attestation_event, register_agent, start_session, pause_session, close_session and 6 more.
Does the openglass-mcp server need sign-in or OAuth?
- Its remote endpoint listed its tools to an anonymous client, without sign-in.
Does the openglass-mcp server send data to outside services?
- 4 tools can carry data out (it sends, posts or publishes, or takes a recipient, channel or URL): invite_counterparty, accept_invite, send_message and send_attestation_event.
Has the openglass-mcp server changed its tools?
- Yes, once since Anlyon Radar started tracking it, most recently on Nov 19, 2026. Every change is on this page and in its Atom feed.