Fidelis Memory — Claude Code MCP server: what it does, how to install
Stores notes on your machine and serves the exact wording back to your agent without LLM rewriting. Helpful when summaries drop critical edge cases or failure conditions.
claude mcp add --env FIDELIS_PORT=<FIDELIS_PORT> --transport stdio fidelis-memory -- uvx fidelis-memoryWhat Fidelis Memory does
The server keeps plain text and Markdown files in ~/.cogito/store on your disk. Most memory integrations condense past context, which often strips out strict boundaries or reasons behind a rollback. Fidelis returns the precise passage you entered, feeding the raw text directly to the assistant during recall.
Running fidelis init registers a background daemon via launchd on macOS or systemd --user on Linux. Retrieval uses local vector embeddings managed by an Ollama instance with the nomic-embed-text model. No external API key is needed.
Without a live Ollama instance, embedding queries will fail immediately. Release 0.3.0rc1 operates strictly on a single machine rather than acting as a networked team database. When you record an update to an existing note, the system retains the previous record alongside its historical status instead of silently overwriting it.
What it helps with
- retrieve verbatim technical conditions across agent sessions
- preserve original incident details without generative summarization
- track note corrections without discarding previous history
- index local markdown files for agent search via Ollama
Example prompts for the agent
- Look up my migration note in Fidelis and quote the retry requirement
- Save the staging rollback details and verification steps into Fidelis
- Check Fidelis memory for earlier review comments on this database schema
How to install Fidelis Memory
Platform CLI
claude mcp add --env FIDELIS_PORT=<FIDELIS_PORT> --transport stdio fidelis-memory -- uvx fidelis-memoryReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from fidelis-memory <command>
Platform CLI
codex mcp add fidelis-memory --env FIDELIS_PORT=<FIDELIS_PORT> -- uvx fidelis-memoryReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from fidelis-memory <command>
Config file
{
"mcpServers": {
"fidelis-memory": {
"command": "uvx",
"args": [
"fidelis-memory"
],
"env": {
"FIDELIS_PORT": "<FIDELIS_PORT>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from fidelis-memory <command>
Config file
{
"servers": {
"fidelis-memory": {
"type": "stdio",
"command": "uvx",
"args": [
"fidelis-memory"
],
"env": {
"FIDELIS_PORT": "<FIDELIS_PORT>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from fidelis-memory <command>
Platform CLI
gemini mcp add -e FIDELIS_PORT=<FIDELIS_PORT> fidelis-memory uvx fidelis-memoryReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from fidelis-memory <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 Fidelis Memory
How do I install Fidelis Memory in Claude Code?
Run claude mcp add fidelis-memory -- uvx fidelis-memory. If you need to specify a port, run claude mcp add --env FIDELIS_PORT=<FIDELIS_PORT> --transport stdio fidelis-memory -- uvx fidelis-memory.
What does Fidelis Memory do?
It saves notes into a local database and returns their unedited passages to the agent. Search operates locally without asking an LLM to restate or condense the text.
What are the system requirements for running it?
The package requires Python 3.10+, macOS or Ubuntu, and an active Ollama process serving nomic-embed-text.
