Crumbtrail MCP server

Read from the package source without running it. The installed server may list more.

• getBug

  Get one distinct bug (by bugId from listDistinctBugs) with its full correlated evidence: front-end and back-end evidence refs, optional db diffs, the representative signal, window, and contributing candidate ids.

• getBugErrorContext

  Get all errors/rejections from a bug with surrounding events for context. Best for understanding what happened around each error.

• getBugEvents

  Get events from a bug report, optionally filtered by kind (clk, con, err, net.req, net.res, key, etc.) or time range. Use limit to control token usage.

• getBugFailedRequests

  Get failed HTTP requests (status >= 400) from a bug report.

• getBugLLMContext

  Get a compact bug context optimized for LLM consumption (small key schema + top errors/requests/navs).

• getBugReport

  Get the full bug report including developer note, URL, summary stats (error count, failed requests, event breakdown). Read this first before diving into events.

• getBugVoiceTranscript

  Get the transcribed voice note from a bug report, if the developer recorded one.

• getCookieChanges

  Get all cookie change events from a session

• getErrorContext

  Get error events with surrounding context events

• getEvents

  Get events from a session, optionally filtered by type or time range

• getEvidence

  Resolve a single piece of evidence by ref from hot-plane artifacts only. ref is a candidate id (e.g. cand_0001 -> candidate + its evidence window), an interactive-element signature (sig -> signatures.json entry + occurrence count), or a request/event id (resolved to the candidate whose anchor references it). Returns a small payload; use getWindow for the raw chronological events. Every response ca

• getFailedRequests

  Get failed HTTP requests (status >= 400)

• getFixContext

  Give a coding agent complete bug context for one recorded session. Returns the fix-context.v1 bundle: ranked candidate causes, the primary evidence window with correlated frontend requests, backend spans, and the exact database rows that changed, plus a redaction-aware environment snapshot and a repro hint. Start here when the user asks you to fix a bug captured with Crumbtrail.

• getFrame

  Get a frame image by timestamp (nearest match)

• getFrameById

  Get a frame image by filename

• getIndex

  Get the index.json summary for a session

• getLatestIssue

  The one-call, zero-friction entry point: finds the most recent finalized session with error-class evidence (errors, failed requests, or critical/high-severity candidates) and returns its complete fix-context.v1 bundle \u2014 ranked candidate causes, correlated primary window, environment snapshot, and repro hint \u2014 in a single call. No sessionId, no listSessions round-trip, no setup: call it w

• getLinkedRequestContext

  Get linked frontend/backend request evidence from index.fullStackRequests for a session/request ID

• getRecurrence

  Get a cross-session recurrence rollup by signature from listDistinctBugs({mode:"cross-session"}). Returns first_seen/last_seen, session_count, release_span, app/tenant labels, and per-session occurrences.

• getRegressionContext

  Compare two recorded sessions of the same flow across releases and return the regression-context.v1 witness verdict. Catches escaped regressions \u2014 behavior changes that fired no error \u2014 and hands back the diverging interaction, the causal window of correlated requests, the exact database rows whose values changed, and a repro hint. Input: { sessionA, sessionB } (ids or paths). Use listSe

• getSessionManifest

  Get the session manifest (manifest.json): metadata, error and failed-request markers, timeline, ranked candidates, and an accessPattern hint. The token-bounded entry point for exploring a recorded session \u2014 start drilldown here, then getWindow for raw events in a time window and getEvidence to resolve one candidate, signature, or request id. Hot-plane only. Every response carries tokenEstimat

• getStorageChanges

  Get all storage change events from a session

• getStorageSnapshot

  Get the initial storage snapshot from a session

• getTranscript

  Get audio transcript events from a session

• getWindow

  Get the cold events within the absolute-ms time window [t0,t1] (the same units as manifest.session.startMs/endMs and a candidate.evidenceWindow.start/end). This is the only tool that reads the cold event stream; it is bounded to the window and capped (default/max 500 events) and reports truncation. Use it after locating a candidate window via getSessionManifest/getEvidence.

• listBugs

  List all bug reports in the queue. Returns report summaries sorted newest-first. Use this as the entry point to triage bugs.

• listDistinctBugs

  List the DISTINCT bugs a session hit, grouped deterministically from the ranked candidates (within-session dedup). With mode:"cross-session", scans finalized sessions and returns recurrence rollups by stable bug signature. Use getBug for one session bug, or getRecurrence(signature) for one recurrence.

• listSessions

  List recorded Crumbtrail sessions \u2014 complete app evidence sessions (clicks when present, console, network, backend spans, database row changes, environment and feature flags) ready to hand to an agent. Use this first to find the sessionId for getFixContext (bug context for one session) or getRegressionContext (cross-release regression witness). Supports app/time/release/build filters.

• locateInteractiveElements

  Find interactive components in a session BY IDENTITY. Returns a deterministic ranked list of {signature, role, label, path, occurrences} from the finalized hot-plane interactive-element map, optionally filtered by a label/text substring or an exact role/tag. Ranked by occurrences desc, then label, then signature. Reads hot-plane artifacts only (redaction-safe). Resolve-only (no live actuation); us

• recallSimilarIssues

  Before diagnosing a bug, ask: have we seen this before? Recalls past issues that RHYME with a session or a free-text description \u2014 not just exact duplicates, but same route/different error, same error/different route, or similar environment/feature-flag state \u2014 ranked by a hybrid of text similarity and structured overlap. Each match carries how it was resolved when known (disposition, ro

• resolveBug

  Mark a bug as resolved.

• resolveSignature

  Resolve a stable component signature to its full interactive-element descriptor for one session: path/selector, tag/role, accessible label/text, occurrence count, first-seen, and interaction affordances (clickable/input). Reads the finalized hot-plane bundle only (redaction-safe; raw masked values are never surfaced). Unknown signature returns an error. Use locateInteractiveElements first to find

• solveContext

  Fuse a described symptom with locally recorded evidence into a fusion.v1 RankedBundle \u2014 the complete, neutral evidence (ranked by relevance) plus a structurally-separate advisory opinion of ranked hypotheses and any evidence gaps. Never a boolean verdict. Pass baselineSession + currentSession to gather evidence via compareSessions; omit both to get an evidence-free bundle describing gaps.

• Code scan8 source files scanned25/25
• –Live reliabilityno gateway calls yet and no remote to proben/a
• –Tool poisoningtools not inspected (local package is not executed); not countedn/a
• Auth qualitylocal package, no credentials required12/15
• Maintenancerepository not readable: repo not found3/15
• Maintainer identityno repository or website to verify2/10

Safety score

65/100

higher than 15% of 802 Databases servers

Tools

33

the typical Databases server has 8

Tools that change things

0%

typical: 10%. Fewer write tools means less an agent can break

Downloads a week

51

more than 36% of 328 Databases servers with a package

Crumbtrail: common questions

Is Crumbtrail MCP server safe?

With care: it is graded C, so read the findings first (65/100). Read the Crumbtrail safety report

How do I install Crumbtrail?

It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.

Does Crumbtrail need an API key?

Not as far as the registry entry and our scan can tell: no credentials are declared or required.

Is Crumbtrail maintained?

The latest release is v0.3.0.

What can I use instead of Crumbtrail?

Servers from other publishers that do the same job: Appwrite MCP server, Magic Cloud MCP server and Rebase MCP server.

How does Crumbtrail compare with other Databases MCP servers?

safety score 65/100 (higher than 15% of 802 Databases servers); tools 33 (the typical Databases server has 8); tools that change things 0% (typical: 10%. Fewer write tools means less an agent can break). Full comparison.