Lunary MCP server

What the model can do

Lunary LLM observability: runs (LLM calls, traces and threads) by type, model, tag and time with status, duration, tokens, cost and feedback, token and cost usage by day and model, prompt templates with their versions, and evaluation datasets; when you ask, can tag, rate and score runs, create prompt templates and versions, publish or delete templates, and create, fill or delete datasets.

Reads

• lunary_list_runsLunary runs: LLM calls, traces or conversation threads, filtered by type, words, model, tag and creation time, with status, duration, tokens, cost…
• lunary_get_runOne Lunary run's metadata: type, name, status, timing, tokens, cost, tags, template, feedback and its trace.
• lunary_usageLunary token and cost usage over the last N days, by model (or by day with daily), with errors and successes: "what did we spend on GPT this week".
• lunary_templatesLunary prompt templates: name, slug, mode, group and their versions (number, draft or live, notes).
• lunary_datasetsLunary evaluation datasets: name, description, how many items, current version and when changed.

Acts, when you ask

These change something in Lunary. The model is told to use them only when you ask for exactly that, and each one is granted by name: per team on a shared connection, by the owner on a personal one.

• lunary_tag_runAdds tags to a Lunary run, keeping the tags it already has.
• lunary_rate_runSets the thumbs up or down feedback (and an optional comment) on a Lunary run, replacing any feedback it had.
• lunary_score_runRecords a named score on a Lunary run (e.g. accuracy 0.95), with an optional comment.
• lunary_create_templateCreates a Lunary prompt template (chat messages, with {{variables}}) under a new slug, as a draft unless publish is true.
• lunary_create_template_versionAdds a new version to a Lunary prompt template.
• lunary_publish_template_versionPublishes a draft Lunary template version, so apps fetching the template's latest version receive it from now on.
• lunary_delete_templatePermanently deletes a Lunary prompt template and all its versions; apps fetching it by slug stop receiving it.
• lunary_create_datasetCreates an empty Lunary evaluation dataset.
• lunary_add_dataset_itemAdds one item (an input, with an optional expected answer and output) to a Lunary dataset.
• lunary_delete_datasetPermanently deletes a Lunary dataset with all its items and versions.

Connect it

First connect Lunary on Dashboard → Connections, then add Memtro to your client once. The Lunary tools appear alongside the tools of everything else you have connected; nothing is configured per connector.

claude plugin marketplace add memtro-org/claude-memtro
claude plugin install memtro@memtro --config api_key=sk-memtro-...

{
  "mcpServers": {
    "memtro": {
      "url": "https://platform.memtro.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk-memtro-..."
      }
    }
  }
}

{
  "mcp": {
    "memtro": {
      "type": "remote",
      "url": "https://platform.memtro.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk-memtro-..."
      }
    }
  }
}

Who may use it

The key identifies the person, and the server offers that person the tools they may use everywhere else in Memtro. A shared Lunary connection is used by the teams an administrator granted it to, tool by tool, on the Permissions page; a personal connection is used by its owner alone, and its tools that act stay off until the owner allows them. See Teams and permissions and Connectors.

Ask it like this

• › What was our LLM spend this week, by model?
• › Which runs got thumbs down yesterday?
• › Which versions of the support reply template are live?
• › How many items are in the regression dataset?
• › Tag that run as needs-review.

Questions

Is this an official Lunary MCP server?

No. It is Memtro's MCP server with Memtro's Lunary connector behind it: the connector reads Lunary through its API with the credentials you connect on Dashboard → Connections, and Lunary publishes nothing here. The same endpoint carries every other system you connect and Memtro's memory tools.

What does it cost?

Nothing on top of Memtro. One price per organisation by its number of users, from $50 a month for up to 10 users after a 7-day free trial, with every connector, the MCP server, jobs and memory included. You bring your own model keys.

Can it change anything in Lunary?

Only when you ask, and only what it has been granted: add tags to a run; give a run thumbs up or down feedback; score a run; create a prompt template; add a version to a prompt template (a draft unless you publish it); publish a draft template version, making it live; delete a prompt template (permanent); create a dataset; add an item to a dataset; delete a dataset (permanent). Every other tool reads, searches or lists. An administrator grants each acting tool by name per team, and a personal connection acts only after its owner allows it.

Which clients work?

Any MCP client that speaks Streamable HTTP: Claude Code, Cursor, Claude Desktop, OpenCode, Codex, Zed, Windsurf, Cline and the Anthropic and OpenAI SDKs. There is no local process to run; the server is remote and authenticated with your key.

Other MCP servers · Engineering and operations

• GitHub MCP server
• AWS CloudWatch MCP server
• Sentry MCP server
• ServiceNow MCP server
• Okta MCP server
• Twilio MCP server