> ## Documentation Index > Fetch the complete documentation index at: https://docs.outerlayer.ai/llms.txt > Use this file to discover all available pages before exploring further. # Compare metrics across two windows > Correlational comparison — not causal attribution: returns fleet-behavior tiles for two EXPLICIT, caller-chosen windows (`a`, `b`; UTC calendar dates), side by side. Neither window defaults — this is for "before vs. after a specific change" (e.g. either side of a context sync from `GET /v1/context/changes`), not a rolling comparison. ## OpenAPI ````yaml /openapi.yaml get /v1/metrics/compare openapi: 3.0.3 info: contact: email: hello@outerlayer.ai url: https://www.outerlayer.ai description: >- The OuterLayer Gateway API lets you ingest coding-agent sessions and record evidence of work programmatically. Most teams reach it through the `outerlayer` CLI, which calls it for them. Call it directly to build your own integration or to automate what the dashboard does. Versioning: every endpoint is prefixed with `/v1/`. Breaking changes ship under a new version prefix (`/v2/`, etc.) with a 90-day-minimum deprecation window; anything additive lands on the current prefix. Vocabulary: the dashboard calls this API's `app` a **factory** — the unit that owns a set of repositories, the agent sessions worked on them, and the evidence they produce. They are the same entity. The API keeps `app` in its paths, headers, and field names because those are a published contract; only the product surface was renamed. title: OuterLayer Gateway API version: '1.0' servers: - description: Production (OuterLayer Cloud) url: https://api.outerlayer.ai - description: Local dev server (self-hosted gateway) url: http://localhost:9418 security: - AppId: [] BearerAuth: [] tags: - description: >- What a CLI login token can read about its owner without naming a factory, such as the factories they can use. name: Account - description: >- Coding-agent session ingest (outerlayer sync) and content-addressed session images. name: Agents - description: >- Evidence artifacts (screenshots, recordings, reports, logs) emitted as proof of a change and anchored to pull requests. name: Artifacts - description: >- Pass/fail outcomes of checks run in your own CI or compute, anchored to pull requests with the run URL as proof. name: Emitted Results - description: >- The acceptance criteria of a work item, recorded as one list. The Criteria tab, its count and the item status read that list. name: Criteria - description: >- What any agent found wrong about a change under review or about a rule it ran on, recorded on a work item. name: Findings - description: >- Add or remove a source's statement that it is working on an issue or pull request, and read what is currently live. name: Work Items - description: >- What the factory's GitHub App installation lets builds do on each connected repository, and whether each default branch requires a pull request. name: Repositories - description: >- A runner host's heartbeat — its CLI version, slots and last request for work — and the factory's list of hosts with a derived status. name: Hosts - description: Query server feature availability. name: Capabilities - description: Per-model LLM pricing data. Public, unauthenticated. name: Pricing - description: >- Create, list, and revoke factory API keys. Plaintext returned exactly once at creation. name: API Keys - description: >- Factory CRUD — a factory is owned by an organization and is the unit every other resource hangs off. Lets a headless agent provision one without the dashboard. name: Apps - description: >- Agent-coding session list and full transcript reads, with actor-privacy controls for machine keys. name: Sessions - description: >- Per-model token spend and fleet-wide agent behavior tiles, including two-window comparisons. name: Metrics - description: Synced-commit history of the app's `.outerlayer/` context tree. name: Context - description: >- Session→PR attribution: which agent sessions produced which pull requests, and what each attributed PR cost. name: PRs - description: Service health checks. name: Health - description: >- Org member and role administration, authenticated with an org-scoped management API key (`olk_…`) minted in the dashboard's settings. name: Org Management - description: OAuth 2.1 discovery metadata for MCP connector clients. name: OAuth paths: /v1/metrics/compare: get: tags: - Metrics summary: Compare metrics across two windows description: >- Correlational comparison — not causal attribution: returns fleet-behavior tiles for two EXPLICIT, caller-chosen windows (`a`, `b`; UTC calendar dates), side by side. Neither window defaults — this is for "before vs. after a specific change" (e.g. either side of a context sync from `GET /v1/context/changes`), not a rolling comparison. operationId: compare-windows parameters: - in: query name: aFrom required: true schema: format: date type: string - in: query name: aTo required: true schema: format: date type: string - in: query name: bFrom required: true schema: format: date type: string - in: query name: bTo required: true schema: format: date type: string responses: '200': content: application/json: schema: properties: data: properties: windowA: properties: from: type: string metrics: properties: activeActors: type: number autoApprovedRate: type: number cleanSessionRate: type: number handsOnRate: type: number modelMix: items: properties: model: type: string sessions: minimum: 0 type: integer required: - model - sessions type: object type: array sessions: type: number toolDenialRate: type: number toolErrorRate: type: number totalCost: type: number unpricedSessions: description: >- Sessions with no stored price in this window, kept out of totalCost. type: number required: - sessions - toolErrorRate - cleanSessionRate - handsOnRate - activeActors - totalCost - unpricedSessions - toolDenialRate - autoApprovedRate - modelMix type: object to: type: string required: - from - to - metrics type: object windowB: properties: from: type: string metrics: properties: activeActors: type: number autoApprovedRate: type: number cleanSessionRate: type: number handsOnRate: type: number modelMix: items: properties: model: type: string sessions: minimum: 0 type: integer required: - model - sessions type: object type: array sessions: type: number toolDenialRate: type: number toolErrorRate: type: number totalCost: type: number unpricedSessions: description: >- Sessions with no stored price in this window, kept out of totalCost. type: number required: - sessions - toolErrorRate - cleanSessionRate - handsOnRate - activeActors - totalCost - unpricedSessions - toolDenialRate - autoApprovedRate - modelMix type: object to: type: string required: - from - to - metrics type: object required: - windowA - windowB type: object required: - data type: object description: Fleet metrics for both windows. '400': content: application/json: schema: properties: error: properties: code: type: string message: type: string required: - code - message type: object required: - error type: object description: Invalid request parameters (validation failed). '401': content: application/json: schema: properties: error: properties: code: type: string message: type: string required: - code - message type: object required: - error type: object description: Missing or invalid API key. components: securitySchemes: AppId: description: Factory id the request is scoped to. in: header name: X-Outerlayer-App-Id type: apiKey BearerAuth: description: API key (sk_outerlayer_*) scheme: bearer type: http ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.