AI Agents Accessible APIsDeveloper platforms and DevOps

Developer platforms and DevOps APIs and MCP servers for AI agents

Source control, CI/CD, monitoring, error tracking, incident management and package registries.

12 developer platforms and devops APIs an AI agent can call: auth method and header, official MCP server (11), free tier (12), docs links. Machine-readable JSON included. Agents: fetch catalog/developer-devops.json instead of parsing this page. Last checked 2026-09-24.

Ratings and usage (★) fetched: .

ServiceWhat an agent can doAuthMCP serverFree tierDetails (with sources)NotesProof (ratings, usage)
CircleCI APIAn agent can trigger pipelines, rerun workflows, fetch logs/test results, and validate CI/CD configuration via REST.API key
Circle-Token: <token>
officialfree plan: 30,000 credits/month (~6,000 build minutes on a small Docker resource class), credits expire monthlyBase URL: https://circleci.com/api/v2
Free plan: 30,000 credits (up to 6,000 build minutes on small Docker), up to 5 active users, per month, no card ↗
Example call: trigger a pipeline on a branch: POST https://circleci.com/api/v2/project/{project-slug}/pipeline ↗
Unit price: Compute billed in credits: Linux Medium = 10 credits per minute; extra credits $15 per 25,000 (about $0.006 per Linux Medium minute) ↗
MCP tools: 22: list_runs, get_run, list_run_workflows, get_workflow, list_workflow_jobs, get_job … ↗
Use the v3 API for new work (v2 still supported). CircleCI's hosted MCP is at https://mcp.circleci.com/v1/mcp (OAuth2 sign-in, 24 tools); the mcp-server-circleci npm package is deprecated.GitHub 94★ CircleCI-Public/mcp-server-circleci
SourceForge 3.0★/2 reviews
Datadog APIAn agent can query metrics, logs, traces, monitors, and dashboards, and manage Datadog configuration via REST.API key
DD-API-KEY: <api_key> plus DD-APPLICATION-KEY: <app_key> for some endpoints
officialfree Infrastructure Monitoring plan: 1-day metric retention, up to 5 hosts; other products (APM, Logs, Security) are paidBase URL: https://api.datadoghq.com
Free plan: Infrastructure free plan: up to 5 hosts, 1-day metric retention ↗
Rate limits: Per-endpoint limits (429 when exceeded); metric data-point submission is not rate limited ↗
Example call: query metric timeseries: GET https://api.datadoghq.com/api/v1/query?from={from}&to={to}&query={query} ↗
MCP tools: 13: search_datadog_logs, analyze_datadog_logs, search_datadog_metrics, get_datadog_metric, search_datadog_monitors, search_datadog_incidents … ↗
Datadog MCP server endpoint is site-specific (app.datadoghq.com, datadoghq.eu, etc.); supports OAuth2, personal/service tokens, or DD-API-KEY/DD-APPLICATION-KEY headers.SourceForge 4.9★/7 reviews
Docker Hub APIAn agent can search, push, pull, tag, and manage container images and repositories on Docker Hub.API key
Authorization: Bearer <token> (obtained via login or a Personal/Organization Access Token)
officialfree personal accounts with rate-limited image pulls; paid Pro/Team/Business tiers raise pull limitsBase URL: https://hub.docker.com
Free plan: Personal (free) account: 200 image pulls per 6 hours (unauthenticated 100 per IP) ↗
Rate limits: Abuse limit of thousands of requests/min per IP; X-RateLimit-* headers, 429 with Retry-After; pulls 200/6h free, 100/6h anonymous ↗
Example call: list tags of a repository: GET https://hub.docker.com/v2/namespaces/{namespace}/repositories/{repository}/tags ↗
MCP tools: 11: search, get_namespaces, list_repositories_by_namespace, get_repository_info, check_repository, create_repository … ↗
API is v2-beta; free/anonymous pulls are rate-limited. Docker's MCP Catalog/Toolkit offers 300+ containerized MCP servers. The hub-mcp server runs locally (stdio or self-run HTTP); no hosted endpoint.GitHub 165★ docker/hub-mcp
GitHub APIAn agent can manage repositories, issues, pull requests, releases, and code search via REST or GraphQL.API key / OAuth
Authorization: Bearer <token>
officialfree with any GitHub account; 5,000 req/hr authenticated, 60/hr unauthenticatedBase URL: https://api.github.com
Free plan: 5,000 requests/hour authenticated; 60/hour unauthenticated, no card ↗
Rate limits: 5,000 req/hour per user token (15,000 on Enterprise Cloud); 60/hour unauthenticated; secondary limits e.g. 100 concurrent, 900 points/min ↗
OAuth scopes: repo, read:org, workflow, gist, user, notifications ↗
Example call: create an issue in a repository: POST https://api.github.com/repos/{owner}/{repo}/issues ↗
MCP tools: 9: get_file_contents, issue_read, issue_write, create_pull_request, merge_pull_request, search_issues … ↗
Rate limits scale with auth type (60/hr unauth, 5,000/hr PAT, higher for GitHub Apps); secondary abuse limits also apply; GraphQL uses a separate points-based limit.GitHub 33.2k★ github/github-mcp-server
Smithery 6,059 uses
SourceForge 4.7★/21 reviews
GitLab APIAn agent can manage projects, issues, merge requests, CI/CD pipelines, and repository files via REST/GraphQL.API key / OAuth
PRIVATE-TOKEN: <personal_access_token>
officialfree tier: full REST API access on GitLab.com Free plan; some features gated to Premium/UltimateBase URL: https://gitlab.com/api/v4
Free plan: GitLab.com Free: 100 authenticated API requests/min burst, 5,000/hour; unauthenticated 60/hour ↗
Rate limits: GitLab.com Free: 100 req/min burst and 5,000 req/hour per user; unauthenticated 60/hour ↗
Example call: create an issue: POST https://gitlab.com/api/v4/projects/{id}/issues ↗
MCP tools: 16: get_project, create_issue, get_issue, list_merge_requests, get_merge_request, get_merge_request_diffs … ↗
GitLab's official MCP server is in beta (GitLab 18.6+) at https://<gitlab-host>/api/v4/mcp with OAuth dynamic client registration, on all tiers incl. Free; no GitLab-maintained multi-language SDK, but community libraries (python-gitlab, node-gitlab) exist.SourceForge 4.4★/14 reviews
Globalping Network Measurement APILets an agent run ping, traceroute, MTR, DNS and HTTP tests from thousands of probes worldwide to check reachability, latency and routing of a host.none
No key required; optional token from dash.globalping.io raises limits: 'Authorization: Bearer <GLOBALPING_TOKEN>'
officialNo key; 250 tests/hour anonymously (up to 50 probes per measurement), 500 tests/hour with a free account token (up to 500 probes); extra tests use credits.Base URL: https://api.globalping.io/v1
Free plan: 250 tests/hour anonymous (50 probes/measurement); 500 tests/hour with free account token, no card ↗
Rate limits: 250 tests/hour anonymous, 500 tests/hour registered; 2 GET requests/s per measurement ↗
Async jobs: POST /v1/measurements returns an id; poll GET /v1/measurements/{id} ↗
Example call: run a ping test from probes in chosen countries: POST https://api.globalping.io/v1/measurements ↗
Unit price: 1 credit per probe test (ping from 5 locations = 5 credits); 250 free tests/hour anonymous, 500 with account; sponsors get 4,000 credits per $1 ↗
MCP tools: 11: ping, traceroute, dns, mtr, http, locations … ↗
Async: POST /v1/measurements, then poll GET /v1/measurements/{id}. Remote MCP (Streamable HTTP) takes OAuth or a Bearer token.—
PagerDuty APIAn agent can create/manage incidents, on-call schedules, escalation policies, and services via REST.API key / OAuth
Authorization: Token token=<API_KEY> (or Authorization: Bearer <oauth_token>)
officialfree plan for up to 5 users: 100 monthly SMS/phone notifications, 1 on-call schedule, 1 escalation policy, API accessBase URL: https://api.pagerduty.com
Free plan: Free plan: up to 5 users, 100 SMS/phone notifications, 1 on-call schedule, 1 escalation policy, per month ↗
Example call: create an incident: POST https://api.pagerduty.com/incidents ↗
MCP tools: 12: browse_incidents, manage_incidents, browse_services, manage_services, browse_schedules, manage_schedules … ↗
The self-hosted pagerduty-mcp-server repo is deprecated and archived; use the hosted MCP https://mcp.pagerduty.com/mcp (EU: https://mcp.eu.pagerduty.com/mcp) with an API token or OAuth.GitHub 34★ PagerDuty/api-schema
GitHub 79★ PagerDuty/pagerduty-mcp-server
SourceForge 4.7★/44 reviews
Postman APIAn agent can manage Postman collections, workspaces, environments, mocks, monitors, and API specs via REST.API key
X-API-Key: <POSTMAN_API_KEY>
officialFree plan: 10,000 Postman API calls/month; rate limits ~300 req/min general, stricter for specific endpointsBase URL: https://api.postman.com
Free plan: 10,000 Postman API calls, 1 user, per month ↗
Rate limits: 300 requests/min per API key; monthly call quota by plan (10,000 on Free) ↗
Example call: list collections: GET https://api.postman.com/collections ↗
Reference docs are at learning.postman.com. Remote MCP modes: /minimal (default), /code, /mcp (full, 100+ tools); US host supports OAuth, EU host mcp.eu.postman.com needs an API key. Docs-search MCP at learning.postman.com/_mcp/server. API host https://api.postman.com (EU: api.eu.postman.com); api.getpostman.com is legacySmithery 3 uses
SourceForge 4.8★/24 reviews
Reflag Management APILets an agent create feature flags, change their targeting for users and companies, and evaluate flags at runtime.API key / OAuth
Authorization: Bearer <REFLAG_API_KEY> (scoped Management API key); remote MCP uses OAuth sign-in
officialFree Starter plan: 1,000 monthly tracked users, unlimited seats, 3 environments; no overage (upgrade required)Base URL: https://app.reflag.com/api
Free plan: 1,000 monthly tracked users, unlimited seats, 3 environments; no overage ↗
Example call: create a feature flag: POST https://app.reflag.com/api/apps/{appId}/flags ↗
Formerly Bucket. Management API key is per app with scopes; the Runtime API uses a separate secret key. Remote MCP over HTTP (mcp-remote for clients without OAuth).—
Sentry APIAn agent can query and triage error/performance issues and manage projects, releases, and org data.API key / OAuth
Authorization: Bearer <auth_token>
officialfree Developer plan: 1 user, 5,000 errors/mo, 5GB logs, 50 replays; includes MCP accessBase URL: https://sentry.io/api/0
Free plan: Developer plan: 1 user, 5,000 errors, 5GB logs, 50 replays, 5M spans, per month ↗
Rate limits: Per caller+endpoint request-rate and concurrency limits (values not published) ↗
Example call: list a project's unresolved issues: GET https://sentry.io/api/0/projects/{organization_id_or_slug}/{project_id_or_slug}/issues/ ↗
MCP tools: 2: search_events, search_issues ↗
API is v0 (public endpoints stable, beta endpoints may change); per-endpoint rate limits via X-Sentry-Rate-Limit-* headers; legacy Basic-auth API keys are deprecated in favor of Bearer auth tokens.Smithery 35 uses
Supabase Management APIAn agent can programmatically create and manage Supabase projects, databases, API keys, and config.API key / OAuth
Authorization: Bearer <access_token>
officialfree plan available; rate limit 120 req/min per user/project (lower for heavier endpoints)Base URL: https://api.supabase.com/v1
Free plan: 2 active projects, 500 MB database, 50,000 MAUs, 5 GB egress; paused after 1 week inactivity, per month ↗
Rate limits: 120 req/min per user per project/org; analytics 30/min; database context 10/min ↗
Example call: create a Supabase project: POST https://api.supabase.com/v1/projects ↗
MCP tools: 19: list_tables, list_extensions, list_migrations, apply_migration, execute_sql, query_logs … ↗
Personal Access Tokens carry full account privileges - treat as highly sensitive. Client SDKs (javascript, python, dart, swift, kotlin) are for the data/auth/storage APIs, not the Management API.—

Other categories