Sleuth MCP server: use GraphQL for DORA metrics
Sleuth tracks deploys and DORA metrics, and it now builds AI tooling too. That tooling touches MCP without serving Sleuth's deployment data.
Checked 3 October 2026 · from Sleuth's help center, the sleuth-io GitHub organization, and the official MCP registry
One of 91 guides to monitoring and developer tools. Each covers what an AI coding agent can get out of the tool through MCP, the protocol that lets the agent query the tool directly.
Summary
Sleuth has not published an MCP server for its DORA metrics as of 3 October 2026. Deploy counts and other metrics are reachable through Sleuth's GraphQL API instead.
Two Sleuth repositories involve MCP, and neither exposes those metrics. sx shares skills and MCP configurations across a team. Genie sits in front of other MCP servers.
Sleuth's help center now covers two products, Sleuth DORA and Sleuth Skills. The MCP material belongs to Sleuth Skills.
How Sleuth works with MCP
Sleuth DORA tracks deployments and computes deployment frequency, lead time for changes, change failure rate and time to restore. DORA stands for DevOps Research and Assessment, the research program that defined these four metrics.
Its main public API is GraphQL at https://app.sleuth.io/graphql. The metricsRecap query returns figures such as numOfDeploys for a team or project over a date range.
Requests carry the organization API key in an authorization: apikey <key> header. Sleuth's documentation stresses that the scheme is apikey, not Bearer.
sx is Sleuth's skill-sharing tool, with 306 stars, an Apache-2.0 license and a last push on 2 October 2026. sx cloud serve exposes a team's vault of skills as an MCP endpoint.
Genie, created on 6 May 2026 and last pushed on 12 May, describes itself as a smart MCP client for agents. It fronts servers such as GitHub's behind a run_query tool.
Every vendor here gets the same four checks: official, maintained, distributed and usable. Sleuth passes 0 of them.
| Check | Result | Evidence |
|---|---|---|
| Official | Fail | As of 3 October 2026 Sleuth documents GraphQL and REST APIs for its DORA data, and has not published an MCP server for it. |
| Maintained | Not applicable | Not applicable to DORA data. sleuth-io/sx, the organization's only repository matching an MCP search, is active but manages AI assets. |
| Distributed | Not applicable | The official MCP registry's only match for "sleuth" on 3 October 2026 was LeadSleuth, an unrelated sales-leads product. |
| Usable | Not applicable | Neither sx nor Genie returns a deploy, a lead time or a change failure rate from Sleuth. |
| Part of the work | What Sleuth has here | Can an agent reach it |
|---|---|---|
| CI & release | Deploy tracking and DORA metrics | Not over MCP. The GraphQL API returns them to an agent with an HTTP tool |
| Production observability | Deploy health from impact sources such as monitoring and incident tools | Not over MCP. Connect the impact sources' own servers instead |
| Agent observability & evals | Sleuth Skills: installations, audit log and AI metrics for shared assets | Not over MCP. The Skills GraphQL API at app.skills.new carries them |
An agent can get Sleuth's numbers, but only by calling the GraphQL API itself. An MCP search for Sleuth leads to sx and Genie, which solve different problems.
What can an AI agent do with Sleuth?
An MCP search for Sleuth returns sx, which holds no deploy data
sx installs skills, rules, hooks and MCP server entries into AI clients from a shared vault. Its MCP endpoint serves that vault to claude.ai or chatgpt.com.
Guard: Check what a repository serves before connecting it. For deploys and DORA metrics, go to the GraphQL API.
GraphQL calls fail with a valid API key
Most HTTP tools default to Authorization: Bearer <token>. Sleuth expects the key after apikey instead.
Guard: Set the header explicitly in the agent's HTTP tool or in a small wrapper script.
A query that worked last month now fails
Sleuth notes that its GraphQL API is still under heavy development and subject to change. It also publishes a list of deprecated parts.
Guard: Keep the queries an agent may run in one wrapper, and check Sleuth's deprecation page when one breaks.
Do you need the MCP server at all?
To rebuild DORA inputs from source systems, connect GitHub's server for merges and Actions runs. Add incident.io or Rootly for failures, and let the agent compute the metrics.
For engineering metrics already computed and served over MCP, Faros AI and LinearB publish servers of their own.
Call Sleuth's GraphQL API for its own metrics. Use source-system servers when the agent has to explain a number.
Sleuth integration setup
There is no server to install. The working route is an HTTP call that any agent with a shell or fetch tool can make.
Deploy count for one team
This mirrors the team metrics example in Sleuth's documentation. Replace the organization and team slugs, which are the URL forms of their names.
curl https://app.sleuth.io/graphql \
-H 'content-type: application/json' \
-H 'authorization: apikey <your-api-key>' \
-d '{"query":"query($org: ID!, $s: DateTime!, $e: DateTime!, $t: [ID]) { organization(orgSlug: $org) { metricsRecap(start: $s, end: $e, filters: {teamSlugs: $t}) { numOfDeploys } } }",
"variables":{"org":"<org-slug>","s":"2026-09-01T00:00:00Z",
"e":"2026-09-30T00:00:00Z","t":["<team-slug>"]}}'What the vendor documents
Sleuth's GraphQL documentation links each example to its in-app GraphiQL explorer. That explorer is the quickest way to find the fields an agent's queries should request.
Should you connect an agent to Sleuth?
Only through the GraphQL API, if your agent can make authenticated HTTP calls. Best for sleuth DORA customers who want metrics in an agent and can maintain a few GraphQL queries.
- Strongest fit
- Only through the GraphQL API, if your agent can make authenticated HTTP calls.
- Main advantage
- Deploy counts and DORA figures come back already computed, scoped by team or project.
- Main weakness
- There is no MCP tool, and the GraphQL API is documented as still changing.
- Operational risk
- Connecting
sxor Genie in the belief that they expose Sleuth's deployment data.
From Oodle
One platform for agent traces and infrastructure
Agent traces usually sit in a different product from the rest of your telemetry, so when a slow database makes an agent slow the symptom is in one tool and the cause is in another. Oodle keeps both in one query surface.
See agent observabilityFrequently asked questions
Does Sleuth have an MCP server?
Not for its DORA data. As of 3 October 2026 Sleuth has not published one, and its MCP-related repositories serve skills and other servers.
What is sleuth-io/sx?
A tool for sharing skills, rules, hooks and MCP configurations across a team. It can serve a skills vault over MCP, but it carries no deploy data.
What is Sleuth's Genie?
An MCP client that fronts other servers behind a query tool, so an agent requests only the fields it needs. Its examples front GitHub and Linear, not Sleuth data.
How can an agent get Sleuth's DORA metrics?
Through the GraphQL API at https://app.sleuth.io/graphql, authenticated with the organization API key. The metricsRecap query returns deploy counts by team or project.
