Erkle team memory — Claude Code MCP server: what it does
Erkle team memory — an MCP server for Claude Code that connects coding agents to shared team memory. Agents look up constraints before editing code and propose new findings.
MIT ★ 1 updated 2026-11-13
claude mcp add --transport http erkle https://erkle.ai/mcpWhat Erkle team memory does
The server connects developer assistants to a hosted registry at https://erkle.ai/mcp where conventions, constraints, and architecture lessons live. Before modifying files, an agent queries stored knowledge scoped to the repo or team, preventing mistakes like repeated bad API patterns. After finishing work, the agent suggests newly learned guidelines back to the base.
All submitted memories require human approval: maintainers must review entries before agents can rely on them. To wire up a repository, the author provides npx @erkle/cli init. This CLI inserts server endpoints into local settings and appends instruction markers to AGENTS.md or CLAUDE.md. The helper preserves unrelated config fields, TOML comments, and aborts if JSON files contain syntax errors.
Initial sign-in relies on a browser popup on the local host. Headless boxes and remote containers lack browsers, requiring a token via erkle keys create --env-file ~/.erkle.env. Generated keys remain read-only unless issued with the --write flag.
What it helps with
- Share technical constraints across multiple developer machines
- Enforce human review on lessons learned by autonomous agents
- Prevent conflicting code patterns across different repositories
- Centralize repository conventions outside individual static files
Example prompts for the agent
- Check our stored team constraints for webhook retries before updating the handler
- Look up billing endpoint conventions for acme/billing
- Propose our new database transaction pattern to Erkle memory
How to install Erkle team memory
Platform CLI: remote server
claude mcp add --transport http erkle https://erkle.ai/mcpPlatform CLI: remote server
codex mcp add erkle-team-memory --url https://erkle.ai/mcpConfig file
{
"mcpServers": {
"erkle-team-memory": {
"url": "https://erkle.ai/mcp"
}
}
}Config file
{
"servers": {
"erkle-team-memory": {
"type": "http",
"url": "https://erkle.ai/mcp"
}
}
}Platform CLI: remote server
gemini mcp add -t http erkle-team-memory https://erkle.ai/mcpStep 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 Erkle team memory
How to install Erkle team memory in Claude Code?
Run claude mcp add --transport http erkle https://erkle.ai/mcp in your command line.
What does Erkle team memory do?
It gives agents access to a shared knowledge layer with constraints, checking team rules before edits and submitting new insights for review.
What is required to run Erkle team memory?
Connection to https://erkle.ai/mcp and account authentication via browser login or a token generated with erkle keys create for headless machines.