Lekta: AI readability audit MCP server
dev.lekta/lektalekta.devstreamable-httpv2026.9.0Scanned
Audits how AI answer engines read a page and grades it A+ to F across four measured layers.
5 tools: 3 change things and 2 only read.
- Reads private data · 0
- Reads text others write · 4
- Can send data out · 4
- 2 findings
Tools listed without sign-in. Remote posture
Anlyon alternative
3 of these tools change something. Run those as governed actions.
Declare what an action changes, and Anlyon puts it under a shared limit, binds every approval to the exact request, and grades the outcome. It works on any HTTPS API you can describe as an action.
Anlyon Radar does not run, proxy or govern this MCP server. A tool your own code calls directly is not routed through Anlyon and is not governed by it.
Import an OpenAPI spec →What the Lekta: AI readability audit MCP server can do
The Lekta: AI readability audit MCP server is published by lekta.dev, a domain the MCP Registry verified. It has 5 tools. It can create or change data (3 tools: lekta_audit, lekta_report and lekta_fix_plan), but cannot delete data. 2 tools only read.
4 tools bring in text other people can write (lekta_audit, lekta_report, lekta_diff and 1 more); 4 tools can send data out of your system (lekta_audit, lekta_report, lekta_fix_plan and 1 more).
The fixed checks matched 2 findings (long-description and annotation-conflict). Each one comes with the text it matched.
Its remote endpoint lists its tools without sign-in, and its tools changed once since Anlyon Radar first saw them.
Last scanned Nov 21, 2026, version 2026.9.0.
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.
5 tools
Changes things 3Claude Code: ask
lekta_auditAudits ONE live page for AEO (Answer Engine Optimization), also called GEO (Generative Engine Optimization): whether ChatGPT, Claude, Perplexity, Google AI Overviews and Bing can r…Reads text others writeCan send data out1 finding
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.
- outboundThe input takes a URL (input.url): a request to a URL can carry data to whoever runs it.
Findings
- Info
The description is 2,912 characters long, over the 2,000 this check looks for.
What this check looks forMatched text
description ends: "...for this account, or the per-target ceiling of 1 audit per minute and 5 per hour."
Description, as the server wrote it Audits ONE live page for AEO (Answer Engine Optimization), also called GEO (Generative Engine Optimization): whether ChatGPT, Claude, Perplexity, Google AI Overviews and Bing can reach it, read it and lift answers from it. Fetches the page the way those engines do (a plain HTTP GET as LektaBot, then a headless-Chromium render pass) and grades it A+ to F across four weighted layers: Access 25% (robots.txt permission for 39 published AI crawler tokens, counting vendor user agents and robots.txt control tokens alike: GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot, PerplexityBot, Googlebot, bingbot, CCBot and more), Indexability 25% (noindex / nosnippet / canonical controls, and how much main content survives with JavaScript off), Answerability 30% (concrete data density, external evidence, semantic section structure, hedging), Recency 20% (a machine-readable and visible last-updated date). Returns Markdown under fixed headings: "## Verdict" (grade, score, per-layer scores, and the URL actually measured after redirects), "## Top issues" (failing and warned checks ranked by points lost, tagged critical/serious/moderate/minor, with evidence), "## Fixes" (the full remediation text for each). Unmeasured layers cap the grade at B and say so. USE THIS for "run an AEO audit", "run a GEO audit", "why is my page not cited in AI answers", "can ChatGPT / Claude / Perplexity fetch and read this page", "check my robots.txt for AI crawlers", and again after every deploy to re-measure. The grade measures eligibility preconditions (access, indexability, extractable answers, dates); it does not observe or predict whether any engine cites the page. DO NOT use it when code will parse the result (lekta_report is the same run as JSON), when an audit already exists and you only want the ranked work list (lekta_fix_plan, which never fetches), or to compare two runs (lekta_diff). It scores one page, not a site: no crawling, no sitemap expansion. COST AND BEHAVIOUR: read-only: the audited site is never modified. It READS AND GRADES robots.txt rather than obeying it as a fetch gate: the audit runs even where LektaBot is disallowed, and the report flags that. A blocking call: a live fetch plus a render pass, cut off at a 75 s hard limit. NOT idempotent: a URL not audited in the last 15 minutes starts a fresh run and spends one slot of the daily MCP quota (free plan default 10 fresh audits per UTC day across 1 hostname; every result prints the count used and the host slot in use), while a repeat inside that window replays the cached measurement for free. An unreachable, blocked or HTTP-error target still returns a report, graded U for "unknown", never F. Only these return an error instead of a grade: a timeout or a server-side engine failure, an exhausted daily quota or host slot, 2 audits already in flight for this account, or the per-target ceiling of 1 audit per minute and 5 per hour.
Annotations the server declares
- title: AEO / GEO audit: grade one page A+ to F
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
lekta_fix_planTurns the most recent STORED audit of a URL into an ordered remediation plan: every failing or warned check as a numbered step, biggest point recovery first, each with its severity…Can send data outNo 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
- outboundThe input takes a URL (input.url): a request to a URL can carry data to whoever runs it.
Description, as the server wrote it Turns the most recent STORED audit of a URL into an ordered remediation plan: every failing or warned check as a numbered step, biggest point recovery first, each with its severity, the points that fixing it is expected to regain, and the full fix text. Reads storage only and never fetches, so the plan is exactly as fresh as the last audit of that URL, and any audit of it counts, whether it was started over MCP or from the lekta.dev panel. USE THIS between lekta_audit and the code change, to decide what to fix and in what order, or to hand a developer or another agent a prioritised AEO/GEO work list. It keeps working when the daily fresh-audit quota is exhausted. DO NOT use it to measure or re-measure a page (lekta_audit does that), and do not call it for a URL that was never audited: it returns an error telling you to run lekta_audit first, not an empty plan. For what changed between two runs use lekta_diff; for the raw structured data use lekta_report. COST AND BEHAVIOUR: free and idempotent: no fetch, no quota, no rate limit; it never starts an audit under any circumstance, and repeat calls return the same plan until a newer audit lands. When the stored audit came from an older engine version the output says so and marks the point estimates unreliable. When it has no failing or warned checks, the tool reports that instead of inventing work.
Annotations the server declares
- title: Ordered path to A+ for an audited page
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
lekta_reportRuns the same page audit as lekta_audit and returns the whole report as one JSON document instead of Markdown, for programs, not people.Reads text others writeCan send data out1 finding
Why it is in this group
- The description's first verb is "run": a word for a change.
- 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.
- outboundThe input takes a URL (input.url): a request to a URL can carry data to whoever runs it.
Findings
- Warn
The server marks "lekta_report" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":false,"openWorldHint":true}; The description's first verb is "run": a word for a change.
Description, as the server wrote it Runs the same page audit as lekta_audit and returns the whole report as one JSON document instead of Markdown, for programs, not people. Top-level fields: engineVersion, url (as requested), finalUrl, fetchedAt, durationMs, targetState, robotsDisallowsLektaBot, score (0-100, null when the target could not be measured), grade, coverage, coverageNote, appliedCaps, movers, warningCount, failCount, and fetchMeta (HTTP status, redirect chain, content type, bytes). layers[] carries each layer definition (name, weight) with its score, and layers[].checks[] every check with id, title, status (pass / fail / warn / info / not_applicable / unmeasured), severity, evidence values and full fix text. Scoring internals are stripped by design: rule source names, reference URLs and threshold rationale are absent; each check keeps only its dated evidence-class tag (SPEC, VENDOR, MEASUREMENT, PEER-REVIEWED or PROVISIONAL). USE THIS when code consumes the result, or when you need a field the Markdown verdict never prints: check ids, per-layer scores, raw evidence values, the redirect chain. DO NOT use it to show a person a verdict (lekta_audit is shorter and ranked by point impact), do not call it after lekta_audit for the same URL expecting new data (it is the same measurement in another shape), and do not reach for it to learn WHY a rule exists; that material is deliberately not in the payload. COST AND BEHAVIOUR: the same fetch, 15-minute cache, quota and limits as lekta_audit, with two differences: the JSON carries no quota line, so call lekta_audit when you need the count used, and nothing here is ranked: checks arrive in layer order, not by point impact.
Annotations the server declares
- title: Full audit report as JSON
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
Only reads 2Show 2 read-only tools
lekta_diffCompares the two most recent scorable audits stored for ONE URL and reports the movement: score and grade before → after, which checks improved, which regressed, with both run time…Reads text others writeCan send data outNo findings
Why it is in this group
- The name "lekta_diff" says "diff": a word for reading.
What it can reach
- untrusted-inputThe description's first sentence says "compare" and "url": it reads web pages, which anyone can write.
- untrusted-inputThe input takes a URL (input.url), and the tool reads it: it reads web pages, which anyone can write.
- outboundThe input takes a URL (input.url): a request to a URL can carry data to whoever runs it.
Claude Code: ask, because it is tagged outbound.
Description, as the server wrote it Compares the two most recent scorable audits stored for ONE URL and reports the movement: score and grade before → after, which checks improved, which regressed, with both run timestamps. This is the progress meter of the fix loop. USE THIS to prove a deploy actually moved the number: lekta_audit → apply the fixes → deploy → lekta_audit again → lekta_diff. DO NOT use it to compare two different URLs or two pages of one site (it takes ONE url and compares that page against its own history), and do not call it when fewer than two scorable audits exist for that URL, which returns an error rather than a partial answer. For the ranked work list use lekta_fix_plan; for the current state of a page use lekta_audit. COST AND BEHAVIOUR: free and idempotent: it reads storage only, no fetch, no quota; it never triggers an audit, so no diff exists until you have run lekta_audit twice yourself. Runs that could not be scored (grade U) and runs where a rule crashed server-side are excluded from the comparison entirely, so a failed re-audit is never reported as a regression; it simply is not there. When the engine version changed between the two runs the output says so, because part of the delta is then ours rather than yours.
Annotations the server declares
- title: Compare the last two audits of a page
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
lekta_my_sitesLists the sites saved on the authenticated Lekta account: URL, latest authoritative grade and score, when it was last audited, and the recheck interval where a schedule is on.Reads text others writeNo findings
Why it is in this group
- The description's first verb is "list": a word for reading.
What it can reach
- untrusted-inputThe description's first sentence says "list" and "url": it reads web pages, which anyone can write.
Description, as the server wrote it Lists the sites saved on the authenticated Lekta account: URL, latest authoritative grade and score, when it was last audited, and the recheck interval where a schedule is on. This is account inventory, not a measurement. PARAMETERS: none; this tool takes no arguments; call it with an empty arguments object. USE THIS first in a session to see which URLs this account already tracks before picking one to pass to lekta_audit, lekta_fix_plan or lekta_diff, or to answer "which of my pages has the worst grade" and "which pages are on a schedule". DO NOT expect a fresh measurement (lekta_audit does that), sites belonging to other accounts, or arbitrary URLs, and do not expect to change anything here: sites are added and removed on the web panel at lekta.dev/en/panel, not over MCP. An account with no saved sites gets an explicit empty-list message, not an error. COST AND BEHAVIOUR: free and idempotent: no fetch to any site, no quota. Each grade shown is the last AUTHORITATIVE (scorable) run for that URL, the same definition the web panel and the badge use; when a newer attempt existed but could not be scored, that is stated on its own line, so the grade is neither hidden nor silently presented as current.
Annotations the server declares
- title: List saved sites on this account
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
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 write4
- 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, 18:52 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
- 3
- Protocol version
- 2025-06-18
- WWW-Authenticate on 401
- not observed
- resource_metadata in WWW-Authenticate
- not observed
- RFC 9728 protected resource metadata
- https://lekta.dev/.well-known/oauth-protected-resource/mcp
- Authorization servers
- not observed
- Legacy HTTP+SSE transport
- no
- Dynamic client registration
- not observed
- Client ID metadata documents
- not observed
- WarnThe whole server
The endpoint served tools/list without credentials, and the list has 3 tools that change things (3 write). Whether a call also works without credentials was not tested: tools are never called.
What this check looks forMatched text
https://lekta.dev/mcp: tools/list returned 5 tools to an anonymous client, including lekta_audit (write), lekta_report (write), lekta_fix_plan (write)
Findings
Matches of fixed checks, with the text they matched.
- Warnlekta_report
The server marks "lekta_report" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":false,"openWorldHint":true}; The description's first verb is "run": a word for a change.
- Infolekta_audit
The description is 2,912 characters long, over the 2,000 this check looks for.
What this check looks forMatched text
description ends: "...for this account, or the per-target ceiling of 1 audit per minute and 5 per hour."
History
Changes to its tool definitions and to its remote endpoint, newest first.
Tools5 changed.~5 changedView diff
- Changed:lekta_auditChanges thingsdescriptioninputSchema
- Changed:lekta_diffOnly readsdescriptioninputSchema
- Changed:lekta_fix_planChanges thingsdescriptioninputSchema
- Changed:lekta_my_sitesOnly readsdescription
- Changed:lekta_reportChanges thingsdescriptioninputSchema
Install
Pinned to version 2026.9.0. Values in angle brackets are yours to fill in.
claude mcp add --transport http lekta https://lekta.dev/mcp --header 'X-API-Key: <your-x-api-key>'claude mcp add --transport http lekta https://lekta.dev/mcp --header 'X-API-Key: <your-x-api-key>'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.
{
"permissions": {
"allow": [
"mcp__lekta__lekta_my_sites"
],
"ask": [
"mcp__lekta__lekta_audit",
"mcp__lekta__lekta_report",
"mcp__lekta__lekta_fix_plan",
"mcp__lekta__lekta_diff"
],
"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 lekta. Review it before you use it.
Capability tags behind 5 rules
- mcp__lekta__lekta_audit untrusted-input, outbound
- mcp__lekta__lekta_report untrusted-input, outbound
- mcp__lekta__lekta_fix_plan outbound
- mcp__lekta__lekta_diff untrusted-input, outbound
- mcp__lekta__lekta_my_sites 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.lekta/lekta/verify?hash=dfaf92e6cba9f4a71a5bc6eb46f6b305faafcf2fbabddfc4ae7c0c015bbeb840" | jq -e '.changed == false'Pinned hash dfaf92e6cba9, from Nov 21, 2026, 13:18 UTC.
Watch
Get told when its tools change: a new tool, a new description, a new class.
https://radar.anlyon.com/servers/dev.lekta/lekta/changes.xmlAn Atom feed. Works in any feed reader or Slack's RSS app.
Badge
[](https://radar.anlyon.com/servers/dev.lekta/lekta)Source: MCP Registry. Synced Nov 21, 2026, 13:17 UTC.
Lekta: AI readability audit MCP server: questions and answers
Answered from the scan above, not from the publisher's description.
Is the Lekta: AI readability audit MCP server safe?
- No scan can promise that. What its definitions show: 5 tools, 3 of which can change, send or delete something, and 2 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 Lekta: AI readability audit MCP server delete data?
- No tool is classed as destructive: going by their definitions, none of them deletes, cancels or revokes anything.
Can the Lekta: AI readability audit MCP server do things that cannot be undone?
- No tool is classed as irreversible: none of them sends, pays, publishes or merges, going by their definitions.
What tools does the Lekta: AI readability audit MCP server have?
- 5 tools: 3 write, 2 read. lekta_audit, lekta_report, lekta_fix_plan, lekta_diff and lekta_my_sites.
Does the Lekta: AI readability audit MCP server need sign-in or OAuth?
- Its remote endpoint listed its tools to an anonymous client, without sign-in.
Does the Lekta: AI readability audit 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): lekta_audit, lekta_report, lekta_fix_plan and lekta_diff.
Has the Lekta: AI readability audit MCP server changed its tools?
- Yes, once since Anlyon Radar started tracking it, most recently on Nov 21, 2026. Every change is on this page and in its Atom feed.