codemem MCP server

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

• activity

  What happened recently: commits pushed and sessions logged, across all machines.

• add_note

  Add a note. kind: decision|howto|resource|issue|idea|note. howto notes are what future sessions get from howto(); decisions explain why something is the way it is.

• add_scan_root

  Register a directory to be scanned for projects on a machine. Remote machines run client/codemem_agent.py which reads its roots from here. Use when a directory is organised enough.

• delete_notewrite action

  Delete a note by id, with its search and embedding rows. Use it to remove a note written by mistake or superseded by a corrected one; the deleted row is returned so it can be written back. Note ids come from search results and the notes list. Deleting is not how you edit a note that is merely out of date — add a newer one instead, because notes are a dated record.

• find_assets

  Find reusable assets. With a query it searches; without, it lists (optionally by kind, tag, project). exclude_maturity="junk,broken,sunset,antiquated" leaves only things worth building on. imports="chromadb" answers "which files/projects use this library" exactly, from parsed import statements.

• handoff

  Record or advance a handoff: one implementation replacing another. One record per name; each call appends a dated stage entry. stage: candidate|shadow|verified|promoted|retired|rolled-back. old/new: asset or project names being replaced/replacing. promotion_class: library|service|agent-facing|pipeline|data-store. report: path to the parity report. note: what happened, failures included. Empty name

• help

  How to use codemem: workflow, tool list, label vocabularies, data sources. Call with no arguments for the overview; a topic (e.g. "maturity", "audience", "vendor", "session") narrows it. Cheap: no DB access.

• howto

  How-to knowledge: publishing to the git host, adding a machine, using codemem, and whatever docs are ingested. Searches howto notes first, then ingested docs. Empty topic lists available howtos.

• link_items

  Link two things. relation: uses|could-reuse|supersedes|derived-from|shares-code-with|deploys-to|related. Give project names, or asset names (from_asset/to_asset) instead.

• list_handoffs

  Handoffs in flight (or at a given stage), newest activity first. Each shows old, new, class, stage, last entry.

• list_projects

  List projects with a one-line summary each. Filter by status, tag, machine (has a location there), audience, maturity (e.g. "authoritative") or exclude_maturity (e.g. "junk,antiquated"). origin: "own" (default) | "vendor" (cloned third-party repos) | "all".

• log_session

  Record what a working session did. Call at the end of meaningful work. summary is required; the rest are free text. resources: tools, APIs, models, other projects used.

• project_brief

  Everything known about one project: description, locations on each machine, reusable assets, recent notes, recent commits, links. Accepts a project name or a working-directory path.

• rate

  Rate a project (or an asset, by giving asset_kind) so future sessions know what is usable and what is not. maturity: authoritative (the one to use) | usable | experimental | antiquated (superseded, prefer newer) | sunset (being retired) | broken | junk (reference only). note: why, and what to use instead. Empty maturity returns the vocabulary and current ratings.

• register_asset

  Record something reusable so it is found next time instead of rebuilt. kind: script|module|function|prompt|skill|mcp-server|docker|service|config|dataset|model|doc|tool|pattern. usage: the one line someone needs to reuse it (command, import, URL). maturity: authoritative|usable|experimental|antiquated|sunset|broken|junk (+ maturity_note why).

• scan_path

  Scan a directory on the SERVER machine now and record the projects found. For other machines use codemem_agent.py. Empty path scans this machine's registered roots.

• search

  Hybrid search (BM25 + embeddings) across projects, assets, notes, commits, locations and docs. kinds: comma list to restrict, e.g. "asset,note". exclude_audience: comma list of project audiences to leave out (e.g. "unrestricted" when working professionally). exclude_maturity: ratings to drop, e.g. "junk,broken,sunset" when you only want things safe to build on. Results carry a maturity field; auth

• stats

  Counts of everything, machines seen, index/embedding coverage, server config.

• update_projectwrite action

  Create or enrich a project. Only supplied fields change. status: active|paused|done|abandoned|archived. audience: unrestricted (default: personal work, unfiltered) | professional | employer (free text, used for filtering). origin: own | vendor (a third-party clone; hidden from search and lists by default). tags: comma list. visibility: private (default, the owner only) | shared (named people) | pu

• verify

  Record that you ran this project (or asset, with asset_kind) today and it worked. Restarts its trust freshness clock (half-life 180 days from verification). note: what you checked. Use after actually exercising the code, not after reading it.

• Code scan17 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 0 days ago15/15
• Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10

Safety score

92/100

higher than 95% of 48,255 servers

Tools

20

the typical server has 7

Tools that change things

10%

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

GitHub stars

2

more than 75% of 28,688 servers

codemem: common questions

Is codemem MCP server safe?

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

How do I install codemem?

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

Does codemem need an API key?

No secret keys are declared. It reads 7 settings from the environment.

Is codemem maintained?

The last commit was 40 days ago (2026-09-25). The latest release is v0.1.0.

What can I use instead of codemem?

Servers from other publishers that do the same job: Deja Vu MCP server, Claudexor MCP server and e2e MCP server.

How does codemem compare with other MCP servers?

safety score 92/100 (higher than 95% of 48,255 servers); tools 20 (the typical server has 7); tools that change things 10% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.