Cogmemory MCP server

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

• add_observation

  Attach a fact/observation to a knowledge graph entity

• add_plan_item

  Add a roadmap/plan item

• analyze_impact

  Analyze the impact of uncommitted changes. Auto-detects changed files via `git diff`, maps them to indexed symbols, and computes the reverse transitive caller closure. Returns impacted symbols with depth and path.

• annotate_symbol

  Attach narrative text to a symbol or execution trace (storage-only — the server never generates the text)

• check_for_updates

  Check if a newer version of cogmemory-mcp is available on npm. Makes a read-only HTTPS GET to registry.npmjs.org (cached 24h). Disable with COGMEMORY_DISABLE_UPDATE_CHECK=1 or config.json disable_update_check=true.

• check_index_coverage

  Report indexed vs. unindexed vs. stale files and per-language breakdowns. Use to understand how much of the workspace the code graph covers.

• cogmemory_status

  Show runtime configuration: package version, schema version, db path, workspace root, scope, index coverage, and subsystem counts. Use this to understand the current state of CogMemory.

• create_entitywrite action

  Add an entity to the knowledge graph (deduped on name+type)

• create_relationwrite action

  Link two entities with a typed relation

• create_specwrite action

  Store a long-form document (PRD, SRS, design doc), optionally linked to a knowledge graph entity

• create_taskwrite action

  Create an actionable task, optionally linked to a plan item

• delete_by_keywrite action

  Delete a context entry by its string key

• delete_by_pathwrite action

  Remove a file from the code graph file_index

• delete_itemwrite action

  Delete a single row by ID from any CogMemory subsystem that uses an integer primary key. For 'context' use delete_by_key; for 'file_index' use delete_by_path.

• end_session

  Close a work session and optionally store a summary

• find_dead_code

  Find symbols with zero inbound structural edges (calls, imports, extends, implements). Excludes exported symbols, configurable entry-point patterns, and test files.

• find_duplicates

  Find duplicate/clone symbol pairs via exact body_hash match and MinHash similarity. Inserts SIMILAR_TO edges for pairs above threshold.

• find_related

  Find semantically-related symbols using structural heuristics: shared callers, shared imports, same file, existing SIMILAR_TO edges. Inserts SEMANTICALLY_RELATED edges.

• generate_codemap

  BFS from an entry symbol through the code graph, returning a bounded subgraph with optional traces and annotations

• get_active_context

  Read the current value of a context key

• get_code_snippet

  Get the source code lines for a symbol by ID or name. Returns the file content between start_line and end_line, with optional context padding.

• get_session_summary

  Recall what happened in a previous session, including its decisions, errors, and changelog entries

• get_spec

  Retrieve a spec by ID or exact title

• index_codebase

  Walk the workspace and extract symbols + edges. Dispatches files to the correct language analyzer automatically via the analyzer registry. Incremental by default (only re-analyzes changed files); use full=true to force complete re-index.

• list_items

  Browse stored entries from any CogMemory subsystem (sessions, decisions, conventions, errors, context, changelog, plan, tasks, entities, relations, observations, specs, symbols, edges, execution_traces, codemap_annotations, file_index). Returns rows with optional filters.

• list_projects

  List all projects known to this CogMemory database, with row counts, last-seen timestamps, and staleness flags. The active project is marked.

• log_change

  Append an entry to the changelog (what happened)

• log_error

  Record an error with its signature, description, and resolution

• prune_projects

  Permanently delete a project and ALL of its memories (decisions, conventions, errors, sessions, code graph, etc.). Requires confirm=true. Irreversible.

• purge_subsystemwrite action

  Remove ALL rows from a single subsystem (and cascaded dependents). Requires confirm=true. Use with care — this is irreversible.

• query_code_graph

  Look up a symbol by name, ID, or file path — returns its callers, callees, and imports (1-hop)

• query_graph

  Multi-hop structural graph query using recursive CTE. Start from a symbol and traverse outbound/inbound/both through edges of specified types up to max_depth. Returns nodes and paths.

• recall

  Unified search across decisions, conventions, errors, and changelog by text, tags, session, or date range

• remember_convention

  Log or update a convention (design token, pattern, style, naming)

• remember_decision

  Log a decision with rationale and optional tags

• rename_project

  Rename a project's display label. The slug (identity) is immutable — renaming never affects memory continuity.

• search_knowledge

  Query the knowledge graph: entities, relations, and observations by name, type, or text

• semantic_code_search

  Semantic/meaning-based code search using TF-IDF over symbol names and source code. Returns symbols ranked by relevance to the natural-language query.

• set_active_context

  Set or update the current active context (upsert by key)

• start_session

  Begin a new work session (returns session ID)

• switch_project

  Switch database, project identity, and workspace root together at runtime. Use when the MCP client pinned a stale --workspace/cwd that does not match the project being worked on (the agent can self-correct without restarting the server). Creates and persists a new project identity if the root has none.

• update_plan_statuswrite action

  Update the status of a plan item

• update_specwrite action

  Update a spec's content, title, or entity link (auto-bumps version)

• update_task_statuswrite action

  Update the status of a task

• Code scan59 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
• Maintenancelast push 9 days ago15/15
• Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10

Safety score

92/100

higher than 93% of 48,243 servers

Tools

44

the typical server has 7

Tools that change things

25%

typical: 0%. Fewer write tools means less an agent can break

Downloads a week

71

more than 53% of 12,306 servers with a package

Grade history

• 2026-09-19restoreB → Ascore 92

Cogmemory: common questions

Is Cogmemory MCP server safe?

Yes, by our scan: it is graded A (92/100). Read the Cogmemory safety report

How do I install Cogmemory?

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

Does Cogmemory need an API key?

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

Is Cogmemory maintained?

The last commit was 30 days ago (2026-10-06). The latest release is v1.15.1.

How does Cogmemory compare with other MCP servers?

safety score 92/100 (higher than 93% of 48,243 servers); tools 44 (the typical server has 7); tools that change things 25% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.