@gitloomhq/mcp
v0.3.0
Published
MCP server for GitLoom — memory for any MCP-speaking agent
Downloads
251
Maintainers
Readme
@gitloomhq/mcp
MCP server for GitLoom Cloud — long-term memory for any agent that speaks the Model Context Protocol.
GITLOOM_API_KEY=gl_live_... npx -y @gitloomhq/mcpThree tools: recall_memory, save_memory and find_skill. Their descriptions and schemas
come from @gitloomhq/sdk, so a
tool behaves identically here, through the SDK, and in the OpenAI or Anthropic
tool formats.
Tools
| Tool | Arguments |
|---|---|
| recall_memory | query; optional filters tags, since, until, time_field, tiers, paths |
| save_memory | fact; optional tags, occurred_at |
| find_skill | task |
since, until and occurred_at take a date (YYYY-MM-DD) or an RFC 3339
time. time_field picks which time since and until bound: occurred (when
it happened, the default), created or updated.
With filters and no query, recall_memory lists every memory they match,
newest first. Each memory it returns starts with the day it happened, when
known:
- [2024-06-01] The user moved to Pune.Configuring a host
The GitLoom CLI writes the config for you, in each host's own shape:
gitloom install claude-code --cloud
gitloom install codex --cloud --write
gitloom install opencode --cloud --namespace support
gitloom install openclaw --cloud
gitloom install hermes --cloudOr by hand:
{
"mcpServers": {
"gitloom": {
"command": "npx",
"args": ["-y", "@gitloomhq/mcp"],
"env": { "GITLOOM_API_KEY": "gl_live_..." }
}
}
}Local memory instead
This server talks to a hosted namespace. To keep the memory on your own machine as a git repository, use the CLI's own server — same protocol, no account:
gitloom init ~/memory
gitloom install claude-code --writeEnvironment
| Variable | Purpose |
|---|---|
| GITLOOM_API_KEY | required — create one at app.gitloom.cloud |
| GITLOOM_NAMESPACE | which namespace to use (default: the account's default) |
| GITLOOM_BASE_URL | override the API host |
