openchronicle-mcp — Claude Code MCP server: setup
Keeps agent memories across sessions with semantic and keyword search. Useful when project decisions must remain available after context changes.
claude mcp add openchronicle-mcp -- uvx openchronicle-mcpWhat openchronicle-mcp does
openchronicle-mcp stores decisions, milestones, and discarded approaches so an agent can retrieve them later. Each project gets its own memory scope, while search combines full-text matching with semantic ranking. It can also inspect a Git repository and turn related commit history into material for long-term memory.
A single ASGI process exposes both a REST API and the MCP HTTP transport. The database stays on your hardware, with versioned schema changes, verified SQLite backups, and optional encrypted copies sent through age and rclone. If the embedding service fails, searches fall back to FTS5 and report the degraded state.
A container needs OC_API_HOST=0.0.0.0; otherwise the application binds only inside the container and the published port cannot reach it. Source installation requires Python 3.14+ with .[mcp,openai]; Claude Code connects over http://127.0.0.1:8000/mcp.
What it helps with
- preserve agent decisions across sessions
- isolate memories by project
- search records by meaning and keywords
- turn Git history into project memory
- create verified SQLite backups
Example prompts for the agent
- Save the SQLite storage decision with the
decisiontag usingmemory_save. - Use
memory_searchto find the project’s earlier storage decisions. - Inspect this repository’s Git history and prepare material for long-term memory.
How to install openchronicle-mcp
Platform CLI
claude mcp add openchronicle-mcp -- uvx openchronicle-mcpuvx runs the command named after the package; if the README names another one, use uvx --from openchronicle-mcp <command>
Platform CLI
codex mcp add openchronicle-mcp -- uvx openchronicle-mcpuvx runs the command named after the package; if the README names another one, use uvx --from openchronicle-mcp <command>
Config file
{
"mcpServers": {
"openchronicle-mcp": {
"command": "uvx",
"args": [
"openchronicle-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from openchronicle-mcp <command>
Config file
{
"servers": {
"openchronicle-mcp": {
"type": "stdio",
"command": "uvx",
"args": [
"openchronicle-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from openchronicle-mcp <command>
Platform CLI
gemini mcp add openchronicle-mcp uvx openchronicle-mcpuvx runs the command named after the package; if the README names another one, use uvx --from openchronicle-mcp <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about openchronicle-mcp
How do I install openchronicle-mcp in Claude Code?
Run claude mcp add openchronicle-mcp -- uvx openchronicle-mcp. This registers the server through uvx.
What does openchronicle-mcp do?
It gives agents persistent, project-scoped memory with combined keyword and semantic retrieval. The same service exposes that data through MCP and REST.
What happens when the embedding provider is unavailable?
Search falls back to FTS5 and exposes the degraded state through /api/v1/health and the MCP health tool. Backfilling resumes after the provider returns.
