← Back to directory
C

Callimachus

Community
Local index & search for your AI coding-agent history.
Category
Dev Tools #60 of 438
Stars
★ 45 Popular
Transport
stdio (local process)
Runtime
Node.js · Prebuilt binary
Credentials
No credential needed
License
AGPL-3.0
Last commit
Tools
21
64FMRS · C

This MCP server provides a comprehensive set of tools for indexing and searching coding agent history, with both read and write operations. It is designed to be local-first, with all data stored locally. The server is part of a larger ecosystem that includes a desktop app and a CLI.

Strongest · Security and permissions 15/20 Weakest · Reliability 10/20

Reliability
10/20
Security and permissions
15/20
Maintenance
12/20
Documentation
14/20
Setup experience
13/20
Why each score
Reliability 10/20
README claims 21 MCP tools (17 read, 4 write) and cargo test entry points, but static review sees no CI workflows or committed tests verifiably covering the server's key paths; error handling and dependency control (Ollama, model downloads) are not evidenced, so a conservative 10.
Security and permissions 15/20
Local-first, keys in OS keychain, no outbound traffic except a user-chosen LLM provider, writes confined to its own memory store; boundary disclosure is good, but confirmation-flow details and SECURITY.md contents are unverified, so not full marks. No red-line risk.
Maintenance 12/20
Clear dual licensing (AGPL-3.0), SECURITY.md, RELEASING.md, changeset versioning, and registry listing; but only 33 stars, 0 open issues gives no signal on response times, and commit/release cadence is unverified — 12.
Documentation 14/20
Extremely thorough README: install, full tool list, CLI examples, and disclosed limitations (macOS-only launchers, Cursor lacking project paths), with links to layered sub-docs; deducted because per-parameter MCP tool docs and troubleshooting depth are not verified in supplied material.
Setup experience 13/20
One-line `npx -y callimachus-mcp` install plus multi-platform build paths, but npx first-run binary download is a supply-chain trust step whose npm/manifest consistency is unverified, and semantic search needs extra setup (model download/Ollama) — 13.

Static review · not runListed 2026-09-08

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesConnects to a database

Best for

  • Developers who want to search their coding agent history
  • Users who work with multiple AI coding tools
  • Local-first workflows prioritizing data privacy

Not for

  • Users who need cloud sync or remote access
  • Users who do not use coding agents (the MCP server offers little value)

Required permissions

  • Needs to access the local index database
  • May execute commands to index coding agent history via the CLI

Risks and side effects

  • Indexing can be resource-intensive, especially during the initial build
  • The local index may contain sensitive data but it never leaves the device

Setup

Before you start

Runtime:Node.js · Prebuilt binary

Run with npx: npx -y callimachus-mcp. You can add it to an MCP-capable client with claude mcp add callimachus -- npx -y callimachus-mcp.

Check that it works

After running claude mcp add callimachus -- npx -y callimachus-mcp, tools like search_threads should appear in your client's tool list; asking "what threads have I worked on recently?" and getting results from the local index confirms the connection.

Troubleshooting

  1. If the MCP server fails to start, ensure Node.js is installed correctly
  2. For issues with npx, check network connectivity and retry
  3. If you use the desktop app, ensure it is running to access the shared index

Things to try

Once connected, you can ask your AI assistant things like:

  • Search my past sessions for discussions about the vector index migration.
  • How did we set up releases before? Give me a cited answer from my history.
  • Record the decision 'use sqlite-vec for KNN' into the project memory.
  • Which sessions touched the file embed/mod.rs?

Tools 21

search_threads read-only
Search coding agent threads across all sources.
search_current_project read-only
Search threads within the repository it runs in.
recent_threads read-only
Get recent threads.
get_thread read-only
Get detailed information about a specific thread.
list_tags read-only
List all available tags.
list_open_todos read-only
List all open TODOs.
get_thread_knowledge read-only
Get distilled knowledge (decisions, gotchas) from a thread.
recall_decisions read-only
Recall past decisions.
Show 13 more tools
recall_gotchas read-only
Recall past gotchas.
find_prior_work read-only
Find historical sessions similar to a task.
project_memory read-only
Get aggregated memory for a project.
ask_history read-only
Ask a question about your history and get a cited answer.
threads_for_file read-only
Find which threads touched a specific file path.
check_decision read-only
Surface settled decisions before re-litigating a proposal.
linked_commits read-only
Get commits that a thread likely produced.
list_snapshots read-only
List session snapshots for a project.
load_snapshot read-only
Restore a saved checkpoint.
complete_todo writes
Close an open TODO.
record_decision writes
Record a decision into a project's memory.
record_gotcha writes
Record a gotcha into a project's memory.
snapshot_session writes
Checkpoint a thread for handoff.

Use cases

Developers who want to search past problems and solutions from their coding agent sessions
Teams maintaining a project knowledge base across AI tools
People who want to resume agent sessions after a context window compaction

Supported clients

Claude Desktop

Listed from the project's documentation, not tested by this site.

Overview

Callimachus is a local-first tool that indexes and searches your history of AI coding agents like Claude Code, Codex, Cursor, and more. It stores all conversations in a local SQLite database, provides hybrid search (keyword and semantic), and comes with an MCP server, a CLI, and a VS Code extension. It is open source under AGPL-3.0 and runs entirely on your machine with no cloud dependency.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 63.3k · Tools 2 Compare with this →

Source revision b3a423a39ea5 Data synced 2026-10-19 Read the FMRS scoring method