Skip to content
verify mcp Beta VerifyMCP is currently in beta. If you notice any issues, get in touch and we’ll put it right.

Engram

NPM · ENGRAM-TOOLS · 2 COMPONENTS

Memory for AI agent teams across tools, sessions, repositories, and teammates.

+3 this week 76 Trust /100

Last scanned (16 hours ago)

Scores higher than 53% of npm servers

Category Memory and Knowledge

Trust breakdown (7 categories)

How this component scores in each security and reliability category. Every signal is checked automatically from public evidence about the published package, including repeated runs of it in an isolated sandbox, and we only credit what we can confirm. How we score → Why this is hard to score →

Supply Chain Security98
  • No malware found by supply-chain analysis.Pass
  • Known CVEs were checked across the 115 of 116 dependencies we could resolve, so this covers what we could see, not the whole tree.Partial
  • No install/post-install scripts declared.Pass
  • Dependency health was assessed across the 115 of 116 dependencies we could resolve, so this covers what we could see, not the whole tree. View diagnostics → Partial
Provenance & Transparency19
Schema Quality & AI Usability62
  • AI-judged instruction clarity (excellent).Pass
  • Context-footprint check failed: tool/resource definitions use about 5661 tokens (~333/item across 17 items; 17 tools + 0 resources), over budget; trim descriptions and params. See how to fix → Fail
  • Usage-examples check failed: none of the tools include examples. See how to fix → Fail
Stability & Change Management100
  • No destabilizing schema changes in the last 30 days.Pass
Tool Coverage100
  • 100% of tools have a non-trivial description (not blank, and not just the tool's name).Pass
  • 100% of tool parameters carry a description.Pass
  • Structured output schemas are declared (100% of tools); any adoption earns full credit.Pass
Tool Safety100
  • No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.Pass
  • All 1 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation.Pass
  • An AI judge read all 18 captured unit(s) of tool text and found none that tries to manipulate the model reading it.Pass
Capabilities100
  • Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.Pass
What we tested

Checks we ran (15)

  • Malware scan
  • Known CVEs
  • Install scripts
  • Dependency health
  • Source repository
  • Provenance
  • License
  • Maintenance
  • Schema quality
  • Stability
  • Tool coverage
  • Injection markers
  • Destructive annotations
  • Judged manipulation
  • MCP protocol

Checks we could not complete (1)

These are gaps on our side, not findings about this package. We only credit what we can confirm, so a category made up entirely of them scores 0.

  • Security disclosure in Provenance & TransparencySecurity-disclosure policy not yet verified: we couldn't inspect the source repository.

Does not apply to this package (2)

  • Endpoint SecurityScored on hosted endpoints only. A package has no live endpoint of its own for us to probe.
  • Transport & ReachabilityScored on hosted endpoints only. A package has no live endpoint of its own for us to probe.

The score does not measure everything. What the score does not measure →

Changelog

Every change we have recorded for this component, newest first. Security-relevant changes are always shown. ▲ marks a change for the better, ▼ a change for the worse; unmarked changes are neutral.

  • 21 Nov 26 0
    • Stability: 0.97 → pass security
  • 20 Nov 26 +1

    No change was recorded against any check on this day. Stability & Change Management went from 93 to 97. That category is still filling its 30-day observation window: 28 days of observed history at the previous scan, 29 at this one. The score rises as the window fills, whether or not the server changes.

  • 17 Nov 26 +1

    No change was recorded against any check on this day. Stability & Change Management went from 83 to 87. That category is still filling its 30-day observation window: 25 days of observed history at the previous scan, 26 at this one. The score rises as the window fills, whether or not the server changes.

  • 15 Nov 26 +1

    No change was recorded against any check on this day. Stability & Change Management went from 77 to 80. That category is still filling its 30-day observation window: 23 days of observed history at the previous scan, 24 at this one. The score rises as the window fills, whether or not the server changes.

  • 13 Nov 26 +1

    No change was recorded against any check on this day. Stability & Change Management went from 70 to 73. That category is still filling its 30-day observation window: 21 days of observed history at the previous scan, 22 at this one. The score rises as the window fills, whether or not the server changes.

  • 11 Nov 26 +1

    No change was recorded against any check on this day. Stability & Change Management went from 63 to 67. That category is still filling its 30-day observation window: 19 days of observed history at the previous scan, 20 at this one. The score rises as the window fills, whether or not the server changes.

  • 9 Nov 26 +1

    No change was recorded against any check on this day. Stability & Change Management went from 57 to 60. That category is still filling its 30-day observation window: 17 days of observed history at the previous scan, 18 at this one. The score rises as the window fills, whether or not the server changes.

  • 7 Nov 26 +1
    • Package version: 0.4.10 → 0.4.11 functional
Diagnostics

Diagnostic detail from the automated scan of this channel: what the scanner observed at each step, so you can see exactly where a check passed or failed. It is informational only and never changes the trust score.

Captured 21 Nov 2026 · Analysed npm/engram-tools@0.4.11

Provenance No attestation

The registry publishes no build provenance for this version, so there is nothing to verify.

Result No attestation
Ecosystem npm

Background: How many MCP packages publish verified provenance →

Dependencies 115 packages
Packages resolved 115
Stale 34
Tree resolution Partial

The dependency tree was only partially resolved, so these counts may be incomplete.

Background: SBOMs and build attestations, explained →

MCP tools · 17 exposed · ~5,337 tokens

The tools this component advertises to a client, with an estimated token cost for each. Expand a tool to see its parameters and schema. The per-tool counts are indicative and are not scored directly; the schema's total context footprint is one signal in Schema Quality & AI Usability. A tool's description is untrusted text the model reads on every call, which is what makes this list a security surface and not just an inventory: how tool poisoning works →

Tool Tokens
engram_close_session ~240

Finish an exact journal at a real completion or handoff. Unlike engram_reflect, this atomically closes the session after saving a supplied final summary or confirming an existing checkpoint. Durable memories and journal history remain; closure ends journal writes without erasing knowledge. Returns JSON with session_id, closed_at, checkpoint_id, already_closed and memories_retained. Get session_id from start/rehydrate/list_sessions. Only this session's member or a human workspace owner may close it; team read access does not grant close authority. A missing checkpoint or failed save leaves it active. Repeating close returns the original receipt without duplicate summaries. Closed journals are read-only for explicit rehydrate and cannot be implicitly reopened. Consumes daily call quota. OAuth: session:write.

NameTypeReqDescription
session_idstringyesExact journal id to close; required. Never pass a repo slug or a natural-language query.
summarystring–Final checkpoint text, at most 6000 characters and no credentials. Omit when reusing an existing checkpoint. Erased journals reject new text; omit to reuse their redacted checkpoint, or start a new s…
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_cross_pollinate ~324

Reuse a lesson solved in another repository before rebuilding a shared pattern. Unlike engram_recall or engram_smarter, this explicitly excludes target_repo from source searches and adds transfer lessons and quality gates. Returns Markdown with ranked memories and source repos from other authenticated workspace repositories, with optional public knowledge. source_repos restricts the search; include_public supplements sparse private results. Records retrieval telemetry without copying knowledge. Consumes workspace daily call quota. OAuth: memory:read.

NameTypeReqDescription
contextstringyesNatural-language task and relevant facts, e.g. migrate our PostgreSQL schema with zero downtime. Never include credentials or raw customer data.
depthstring–Default focused uses a small retrieval budget; deep expands it when focused results are insufficient. An explicit limit overrides the default budget.
include_publicboolean–Default true: supplement private results with public knowledge when needed. False searches only other private workspace repositories.
limitnumber–Maximum retrieved memories, 1–25. Omitted defaults: focused=5, deep=12; explicit values override these defaults.
session_idstring–Exact journal id from start/rehydrate/list_sessions. Disambiguates parallel work; writing requires your active session. Rehydrate may read shared or closed history.
source_reposarray–Restrict source repositories to these slugs; omit or pass an empty array to search other repositories.
target_repostringyesDestination repository slug for the current work; memories from this repo are excluded from the source search.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_delete_memory ~313

Preview or permanently erase one private memory and its explicit derived patterns from the live database. Unlike engram_retract_memory, execution cannot be undone. Requires a human workspace OWNER API key; OAuth agents and member keys cannot erase. Does not delete independent memories or unrelated sessions. Default dry_run=true returns counts, derived IDs, affected session IDs (first 50), scope and confirmation. Inspect it first: execution redacts the entire free-text history/checkpoints/task/context of referencing sessions, plus associated outcomes, links, jobs and linked evidence; shared entities remain if still referenced. Execute only the reviewed ID using dry_run=false and confirmation="delete:<memory_id>". Returns erased_at/counts; repeated execution returns the original minimal ID/count receipt without erased text (a one-way content fingerprint blocks exact store replays). In-flight background writes and exact store retries cannot repopulate erased knowledge. Historical unlinked paraphrases, already delivered client context, application/provider logs, provider copies and backups are outside this live-database operation. Public-bank records are excluded. Consumes daily call quota.

NameTypeReqDescription
confirmationstring–For execution only: copy delete:<exact memory_id> from the preview. Omit for preview; never substitute a repo name.
dry_runboolean–Default true previews impact without changing knowledge. False executes only with the exact confirmation string and human OWNER authority.
memory_idstringyesExact id from a recalled memory; do not substitute a repository name, URL, or the memory text.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_get_context ~232

Audit repository knowledge by category: conventions, gotchas, patterns, solutions and architecture. Unlike engram_recall or engram_smarter, this searches each category separately and returns five sample arrays for coverage review, rather than one task-ranked list. Returns JSON with repo/file/total and up to 21 entries; this is a bounded search, not a full repository-memory export. repo is required; file/task refine search without reading files. Only when all five categories have no repo matches does the search broaden to workspace/public memory. Empty results are valid. Does not edit knowledge. Consumes workspace daily call quota. OAuth: memory:read.

NameTypeReqDescription
filestring–File path/name used only as a search hint, e.g. src/auth.ts. Engram does not read or modify local files.
repostringyesRepository slug to inspect, e.g. billing-api. If no repo-specific memories match, the server broadens to workspace/public knowledge.
taskstring–Concrete current goal or decision, e.g. diagnose checkout retries; used to focus context and session recovery.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_list_sessions ~324

Find an exact session_id before recovery, closing or a handoff. Unlike engram_rehydrate, this lists bounded metadata without replaying journal content or changing a session. Returns JSON sessions with member/client attribution, status, activity and checkpoint availability, plus next_cursor. Default scope=mine lists this member's sessions; scope=workspace explicitly includes teammates in the authenticated workspace. repo/status/agent_tool/member_id filter results; a teammate member_id requires workspace scope. limit defaults to 20, maximum 50. Pass next_cursor unchanged to continue newest-started-first pagination. Empty results are valid. Consumes daily call quota. OAuth: memory:read.

NameTypeReqDescription
agent_toolstring–Calling client name for attribution, e.g. codex, claude-code, cursor, or gemini. Omit if unknown.
cursorstring–Opaque next_cursor from the prior response. Keep filters unchanged; omit for the first page.
limitinteger–Page size, default 20; from 1 to 50.
member_idstring–Member id to filter; another member requires scope=workspace. This is a filter, not authorization.
repostring–Repository slug, e.g. engram or billing-api, within the authenticated workspace. This is a memory filter, not a filesystem path or workspace selector.
scopestring–Default mine lists your sessions across clients; workspace includes shared teammate session metadata.
statusstring–Filter journal state; omit to include both active and closed sessions.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_proof ~88

Audit adoption before a rollout or when checking whether the workspace is using memory and reporting outcomes. Takes no arguments. Returns JSON with memory-bank breakdowns, learning-loop activity, proof points and a benchmark checklist. Does not change knowledge or run a benchmark. Use engram_setup for onboarding and engram_recall for task knowledge. Consumes workspace daily call quota. OAuth: memory:read.

Input schema present but exposes no named parameters.

NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_recall ~385

Answer one concrete question from authenticated workspace/public memory before deciding or debugging. Unlike engram_smarter, this retrieves question-specific evidence without task checklists; unlike engram_get_context, it returns one ranked memory list instead of separate category samples. Use engram_rehydrate to recover an interrupted session. Returns JSON with ranked memories, ids, trust/conflict signals and retrieval hints. Example: {"repo":"engram","query":"What did we learn about Netlify SSR costs?"}. No matches returns count=0; deepen or rephrase rather than assume the answer. repo/type/tags narrow the question; limit overrides the depth budget. Records retrieval telemetry without editing knowledge. Consumes workspace daily call quota; stop on quota errors and retry only transient failures. OAuth: memory:read.

NameTypeReqDescription
depthstring–Default focused uses a small retrieval budget; deep expands it when focused results are insufficient. An explicit limit overrides the default budget.
limitnumber–Maximum retrieved memories, 1–20. Omitted defaults: focused=5, deep=12; explicit values override these defaults.
querystringyesSpecific decision, error or question to search, e.g. how are checkout retries deduplicated?
repostring–Repository slug, e.g. engram or billing-api, within the authenticated workspace. This is a memory filter, not a filesystem path or workspace selector.
session_idstring–Exact journal id from start/rehydrate/list_sessions. Disambiguates parallel work; writing requires your active session. Rehydrate may read shared or closed history.
tagsarray–Topic labels for storing or filtering memory; omit when no useful labels apply.
typestring–Knowledge category: team rule, reusable pattern, pitfall, proven fix, design decision, or library/version quirk.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_reflect ~431

Checkpoint a milestone or handoff for later recovery with engram_rehydrate. Supply at least one of summary, learnings, used_memories or feedback. Returns Markdown with stored/skipped counts, outcomes and next actions. Use engram_store for one lesson and engram_report_outcome for one outcome. store=false skips learning storage only: the journal and supplied outcomes/feedback can still be written. This checkpoint leaves the session open. Retrying may duplicate writes; verify prior success first. Consumes workspace daily call quota and memory quota for saved lessons. OAuth: session:write, memory:write and outcome:write.

NameTypeReqDescription
agent_toolstring–Calling client name for attribution, e.g. codex, claude-code, cursor, or gemini. Omit if unknown.
feedbackobject–Optional product feedback or feedback-pulse dismissal; separate from outcomes about individual memories.
learningsarray–Durable private workspace lessons to persist when store=true; only the first 8 are processed. Low-quality entries may be skipped. The summary remains a journal checkpoint.
repostring–Repository slug, e.g. engram or billing-api, within the authenticated workspace. This is a memory filter, not a filesystem path or workspace selector.
session_idstring–Exact journal id from start/rehydrate/list_sessions. Disambiguates parallel work; writing requires your active session. Rehydrate may read shared or closed history.
storeboolean–Default true saves accepted durable learnings; false skips learning storage only. Session checkpoints, supplied outcomes and feedback are still recorded.
summarystring–Milestone/handoff summary: what changed, evidence, decisions and remaining work. Required unless learnings, outcomes or feedback are supplied.
taskstring–Concrete current goal or decision, e.g. diagnose checkout retries; used to focus context and session recovery.
used_memoriesarray–Outcomes for memories actually used during this milestone; only the first 20 are processed. Omit when none were applied.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_rehydrate ~451

Recover interrupted work after compaction, resume or lost context. Returns Markdown checkpoints, decisions, gotchas, open loops and recent events. Unlike engram_start, this restores the prior journal when available; unlike engram_recall or engram_smarter, it recovers session history rather than only searching durable knowledge. session_id selects an accessible workspace session. Closed or teammate journals are read-only history, never reopened. Automatic recovery requires one active journal for this member, repo/project and authenticated OAuth connection or supplied agent_tool within the recovery window. Multiple matches return 409: list sessions and choose an id. since filters events after selection. If none matches, loads durable task memory and opens a journal. Repeating active recovery adds events; it does not erase history. Consumes workspace daily call quota. OAuth: memory:read and session:write.

NameTypeReqDescription
agent_toolstring–Calling client name for attribution, e.g. codex, claude-code, cursor, or gemini. Omit if unknown.
depthstring–Default standard: up to 30 events/4 summaries; brief uses 12/2, full 60/8. This controls journal detail, not retrieval depth.
projectstring–Product/project name for session context. Start/setup use it as the repository hint when repo is omitted.
reasonstring–Recovery trigger, e.g. context compaction or resumed after interruption; recorded in the journal.
repostring–Repository slug, e.g. engram or billing-api, within the authenticated workspace. This is a memory filter, not a filesystem path or workspace selector.
session_idstring–Exact journal id from start/rehydrate/list_sessions. Disambiguates parallel work; writing requires your active session. Rehydrate may read shared or closed history.
sincestring–Optional ISO 8601 lower bound for journal events, e.g. 2026-10-16T10:00:00Z; omit for normal recent-session recovery.
taskstring–Concrete current goal or decision, e.g. diagnose checkout retries; used to focus context and session recovery.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_report_outcome ~283

Record whether a recalled memory helped; returns a confirmation. Use its exact memory_id and evidence in detail. success=false may include failure_reason; outdated marks obsolete knowledge, while success raises confidence. Only accessible workspace/public memories may be reported. Repeating adds outcomes and changes confidence; do not blindly retry. Use engram_store with supersedes_memory_ids for a verified replacement, or engram_reflect to batch milestone outcomes. Consumes workspace daily call quota. OAuth: outcome:write.

NameTypeReqDescription
detailstring–Brief evidence of how the memory changed the decision, or why it failed. Exclude secrets.
failure_reasonstring–Why an unsuccessful memory failed; outdated marks obsolete knowledge. Omit for a successful use.
memory_idstringyesExact id from a recalled memory; do not substitute a repository name, URL, or the memory text.
repostring–Repository slug, e.g. engram or billing-api, within the authenticated workspace. This is a memory filter, not a filesystem path or workspace selector.
session_idstring–Session where this memory was retrieved; used for outcome attribution, not to reopen or modify that journal. May be closed.
successbooleanyesWhether this memory helped the actual task: true for a useful application, false for a misleading or unusable result.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_restore_memory ~161

Make an explicitly retracted private memory available again. Unlike engram_store, this restores the original ID/content; storing matching retired content returns 409 rather than silently reviving it. Returns JSON memory_id, ACTIVE status and already_active. Retained source-dependent history becomes available when no other withdrawn sources remain. Only the creator or a human workspace OWNER may restore. Memory quota still applies. A confirmed replacement or SUPERSEDED state requires the separate replacement-review workflow; permanently erased IDs cannot be restored. An already active record is an idempotent no-op. Consumes daily call quota. OAuth: memory:write.

NameTypeReqDescription
memory_idstringyesExact id from a recalled memory; do not substitute a repository name, URL, or the memory text.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_retract_memory ~229

Withdraw an exact private memory that should no longer guide agents. Unlike engram_store with supersedes_memory_ids, this needs no replacement and keeps the content for explicit restoration. Unlike engram_delete_memory, it is reversible. Returns JSON memory_id, RETRACTED status and already_retracted. Get memory_id from recall/context; provide a non-sensitive reason, 1–1000 characters. Only the creator or a human workspace OWNER may withdraw it; team read access is not write authority. Current retrieval excludes it and source-dependent patterns. Recovery suppresses free text in referencing journals rather than replay withdrawn guidance. Start a clean session for new work or restore the source. Repeating is idempotent. Public-bank and superseded records are excluded. Consumes daily call quota. OAuth: memory:write.

NameTypeReqDescription
memory_idstringyesExact id from a recalled memory; do not substitute a repository name, URL, or the memory text.
reasonstringyesWhy this knowledge should stop guiding agents, 1–1000 characters. No secrets or copied sensitive content.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_setup ~318

Explain onboarding and inspect the authenticated workspace when an agent needs setup guidance. Returns one JSON response with plan/counts/limits, tool groups, activation checklist and suggested calls; no pagination or installation. Hints personalize examples: project names the work; repo scopes activation guidance, falling back to project when omitted. Does not configure the machine, open a working session or edit knowledge. Use engram_start to begin work, engram_rehydrate to recover it and engram_proof to audit adoption. Consumes workspace daily call quota; exhaustion requires reset or a plan change. OAuth: memory:read.

NameTypeReqDescription
agent_toolstring–Calling client name for attribution, e.g. codex, claude-code, cursor, or gemini. Omit if unknown.
marketstring–Business/domain context, e.g. fintech or education; not a geographic access setting.
projectstring–Product/project name for session context. Start/setup use it as the repository hint when repo is omitted.
repostring–Repository slug, e.g. engram or billing-api, within the authenticated workspace. This is a memory filter, not a filesystem path or workspace selector.
rolestring–Agent responsibility used in task context, e.g. backend engineer. Start/setup default to implementation agent.
stackarray–Project technologies used to focus context; omit when unknown.
taskstring–Concrete current goal or decision, e.g. diagnose checkout retries; used to focus context and session recovery.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_skill ~160

Find a procedural checklist before a repeatable task. Omit skill or pass an empty string to list available names; otherwise load a name such as code-review or skill:code-review. Returns JSON with available_skills/usage for listing or skill/count/steps/combined for a guide. Unknown names return count=0 and a message. Reads workspace/public knowledge without installing or executing anything. Use engram_recall for a project fact and engram_smarter for task-based guide selection. Consumes workspace daily call quota. OAuth: memory:read.

NameTypeReqDescription
skillstring–Checklist name, e.g. code-review, security or skill:deployment. Omit or pass an empty string to list available names; unknown names return count=0.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_smarter ~396

Brief a materially different task within the same session. Unlike engram_recall, this combines goal-focused memories with optional procedural checklists; unlike engram_get_context, it ranks evidence for the task instead of returning category samples. Returns Markdown with ranked workspace/public memory and retrieval hints. Use engram_start to begin work and engram_rehydrate for lost context. Example: {"repo":"engram","context":"Improve public-page delivery on Netlify while preserving localized SEO and onboarding."}. context states the whole goal; repo/file focus retrieval without reading local files. limit overrides depth; include_skills=false suppresses checklists. Records retrieval/session telemetry without editing knowledge. Consumes workspace daily call quota. OAuth: memory:read.

NameTypeReqDescription
contextstringyesNatural-language task and relevant facts, e.g. migrate our PostgreSQL schema with zero downtime. Never include credentials or raw customer data.
depthstring–Default focused uses a small retrieval budget; deep expands it when focused results are insufficient. An explicit limit overrides the default budget.
filestring–File path/name used only as a search hint, e.g. src/auth.ts. Engram does not read or modify local files.
include_skillsboolean–Include relevant procedural checklists. If omitted, the server infers this from task/context keywords; false suppresses checklists.
limitnumber–Maximum retrieved memories, 1–30. Omitted defaults: focused=5, deep=12; explicit values override these defaults.
repostring–Repository slug, e.g. engram or billing-api, within the authenticated workspace. This is a memory filter, not a filesystem path or workspace selector.
session_idstring–Exact journal id from start/rehydrate/list_sessions. Disambiguates parallel work; writing requires your active session. Rehydrate may read shared or closed history.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_start ~585

Begin meaningful work in a repository with a Markdown briefing: session id, ranked private/public memory, workspace guidance and optional checklists. Unlike engram_smarter, this includes workspace onboarding/activation guidance; unlike engram_rehydrate, it prepares a starting briefing instead of reconstructing a prior journal. Use once when beginning work; choose engram_smarter for a task switch and engram_rehydrate for recovery. Supply repo and task when known; context overrides the generated project/market/role/stack/task query. Reuses one eligible active journal; new_session=true deliberately creates a separate journal. Supply session_id for parallel work. Ambiguous automatic selection returns 409; list sessions and choose an exact id. Records a session event; repeating may add events or open another session. Consumes workspace daily call quota; quota exhaustion requires reset or a plan change, not immediate retries. OAuth: memory:read and session:write.

NameTypeReqDescription
agent_toolstring–Calling client name for attribution, e.g. codex, claude-code, cursor, or gemini. Omit if unknown.
contextstring–Optional full task context; overrides the query generated from project, role, market, stack and task.
depthstring–Default focused uses a small retrieval budget; deep expands it when focused results are insufficient. An explicit limit overrides the default budget.
include_skillsboolean–Include relevant procedural checklists. If omitted, the server infers this from task/context keywords; false suppresses checklists.
limitnumber–Maximum retrieved memories, 1–30. Omitted defaults: focused=8, deep=16; explicit values override these defaults.
marketstring–Business/domain context, e.g. fintech or education; not a geographic access setting.
new_sessionboolean–Default false reuses one eligible active journal. True creates a separate journal for parallel work; retries can create another session. Keep the returned session_id.
projectstring–Product/project name for session context. Start/setup use it as the repository hint when repo is omitted.
repostring–Repository slug, e.g. engram or billing-api, within the authenticated workspace. This is a memory filter, not a filesystem path or workspace selector.
rolestring–Agent responsibility used in task context, e.g. backend engineer. Start/setup default to implementation agent.
session_idstring–Exact journal id from start/rehydrate/list_sessions. Disambiguates parallel work; writing requires your active session. Rehydrate may read shared or closed history.
stackarray–Project technologies used to focus context; omit when unknown.
taskstring–Concrete current goal or decision, e.g. diagnose checkout retries; used to focus context and session recovery.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

engram_store ~417

Save one durable lesson privately for future workspace agents; returns a confirmation. Use engram_reflect for a milestone with several lessons/outcomes and engram_report_outcome to assess an existing memory. To correct knowledge, supply the complete replacement in content and old ids in supersedes_memory_ids; change_reason explains the replacement. Old entries retain history; this is not permanent deletion. Similar content may deduplicate and boost confidence, so retries can have effects. Consumes memory/call quota; rejects secrets and low-quality content. OAuth: memory:write.

NameTypeReqDescription
agent_toolstring–Calling client name for attribution, e.g. codex, claude-code, cursor, or gemini. Omit if unknown.
change_reasonstring–Why the supplied memories are superseded, up to 1,000 characters; used with supersedes_memory_ids.
contentstringyesOne durable, self-contained lesson, up to 20,000 characters. State the decision/fix and why it matters; exclude secrets and transient debug logs.
contextobject–Non-sensitive supporting metadata, e.g. {language: "typescript", component: "auth"}; never credentials or raw customer data.
repostring–Repository slug, e.g. engram or billing-api, within the authenticated workspace. This is a memory filter, not a filesystem path or workspace selector.
session_idstring–Exact journal id from start/rehydrate/list_sessions. Disambiguates parallel work; writing requires your active session. Rehydrate may read shared or closed history.
supersedes_memory_idsarray–Up to 20 accessible memory ids explicitly replaced by this lesson; omit for an ordinary new memory. Old entries retain their history.
tagsarray–Topic labels for storing or filtering memory; omit when no useful labels apply.
typestringyesKnowledge category: team rule, reusable pattern, pitfall, proven fix, design decision, or library/version quirk.
NameTypeReqDescription
formatstringyesPayload encoding: Markdown context/checkpoint, serialized JSON data, or plain confirmation text.
textstringyesThe same bounded payload as content[0].text; interpret it using format. JSON text contains the endpoint-specific fields described by this tool.
truncatedbooleanyesTrue when the payload exceeded 120,000 characters and was cut with a notice. Narrow the request; truncated JSON is not parseable.

No examples provided.

Install

How do I install the Engram MCP server?

Engram runs locally as an npm package, launched with npx -y engram-tools mcp. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

npm · engram-tools

# add to Claude Code
claude mcp add tools-engram-engram -- npx -y engram-tools mcp
Common questions

What is the Engram MCP server?

Engram is an MCP server listed in the public MCP registry as tools.engram/engram. Memory for AI agent teams across tools, sessions, repositories, and teammates. This page covers its npm package (engram-tools).

Is the Engram MCP server safe to use?

Engram scores 76 out of 100 on VerifyMCP. It declares no install or post-install scripts. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

What tools does the Engram MCP server expose?

Engram exposes 17 tools: engram_start, engram_rehydrate, engram_recall, engram_store, engram_report_outcome, and 12 more. Their descriptions and schemas cost roughly 5,337 tokens of context every time the server is loaded.

Is the Engram MCP server still maintained?

Engram is still listed as active in the MCP registry. We last reached this channel on 21 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

What licence is the Engram MCP server under?

Engram declares the MIT licence, which is OSI-approved. That covers the source only, and says nothing about the cost of any service it calls.

What did VerifyMCP test on the Engram MCP server?

VerifyMCP ran 15 checks on Engram, covering Malware scan, Known CVEs, Install scripts, Dependency health and 11 more. One item could not be completed, a gap on our side rather than a finding about Engram: Security disclosure. Endpoint Security and Transport & Reachability do not apply to an npm package, so they were not tested. The score does not measure everything, and the scoring documentation lists what it leaves out.