--- title: "Use AI for more of your work. Keep your standards." description: "Predictable public entry points for the EvalGate API, SDKs, CLI, MCP servers, agent skills, and authentication contract." canonical: "https://www.evalgate.com/developers" last-updated: "2026-10-21" --- # Use AI for more of your work. Keep your standards. Public developer hub Connect your repo and EvalGate finds the AI in your code: model calls, prompts, agents, and tools. Test what they do, catch regressions before release, and keep the proof. It is for business and product leaders, ml leads, and coding agents that need to evaluate ai behavior against requirements. [Start with the free Developer plan](https://www.evalgate.com/onboarding) [Machine-readable index](https://www.evalgate.com/developers.md) Coding agent quick start · Experimental `npx skills add evalgate/skills` [Canonical public source](https://github.com/evalgate/skills) [First-party discovery index](https://www.evalgate.com/.well-known/agent-skills/index.json) ## Which agent should I start with? Start with public Skills and a read-only working theory. Never use a different agent's apparent pass as release evidence. Product MCP reads require an approved organization-scoped credential. ### Claude Code Investigate AI behavior and find the first testable gap `npx @evalgate/sdk understand --format json` ### Codex Review a repository change and propose a regression gate `npx @evalgate/sdk init --local --format json` ### Cursor Inspect local readiness before changing code `npx @evalgate/sdk status --json` Preview Skills with `npx skills add evalgate/skills --list`, install with `npx skills add evalgate/skills`, and inspect available CLI commands using `npx @evalgate/sdk capabilities --format json`. Only apply, connect, or publish after reviewing the evidence and scopes. [EvalGate MCP on Smithery ↗](https://smithery.ai/servers/evalgate/evalgate) [Public documentation MCP ↗](https://www.evalgate.com/api/mcp/docs) [Agent setup and access boundaries →](https://www.evalgate.com/docs/guides/agent-discovery) Choose a surface ## One product contract, multiple ways to work. Discovery documents are public. Customer data and billable product actions remain behind organization-scoped credentials. ### REST API and OpenAPI Inspect the OpenAPI 3.2 contract, typed requests, scoped authentication, errors, and rate-limit conventions. [Open OpenAPI contract](https://www.evalgate.com/openapi.json) ### TypeScript and Python SDKs Use the published packages for evaluations, traces, regression gates, repository workflows, and automation-safe clients. [Explore SDKs](https://www.evalgate.com/sdk) ### CLI automation Initialize a project, apply reviewed setup, manage baselines, and produce CI-native or JSON gate results. [Read the CLI guide](https://www.evalgate.com/docs/sdk/cli) ### MCP servers Discover public documentation and authenticated read-only product evidence through Streamable HTTP MCP; apply the canonical Skill decision contract separately. [Configure MCP](https://www.evalgate.com/docs/platform/mcp-integration) ### Agent skills and plugin Give coding agents a behavioral-change decision framework plus bounded setup, trace, repository, MCP, and regression workflows. [Browse the public skill bundle](https://github.com/evalgate/skills) ### Authentication Start with GitHub or Google sign-in, then issue an attributable scoped API key for CLI and agent execution. [Read the auth walkthrough](https://www.evalgate.com/auth.md) ## Sandbox and test mode Try the API without touching production data. Neither option needs an account, reads customer data or creates a charge. ### Hosted test mode `GET` returns the contract and a runnable example; `POST` the example to run deterministic assertions on synthetic text. No authentication, saved results, provider calls or charges. Synthetic supplied text only; does not exercise hosted CRUD, providers or payments. Normal access/API-usage metadata and rate-limit counters apply. `https://www.evalgate.com/api/demo/custom-eval`[Test mode guide](https://www.evalgate.com/docs/platform/api-test-mode) ### Isolated local API sandbox Run the real API against a disposable PostgreSQL database with synthetic credentials, then reset it. Outbound network, provider, email and billing calls are blocked. Requires an EvalGate checkout and Docker. `node scripts/sandbox.mjs start`[Sandbox setup guide](https://www.evalgate.com/sandbox.md) ## Need a deterministic first check? Preview product understanding and repository setup, establish a reviewed baseline, then run a gate. An offline pass proves only the deterministic evidence that actually ran. `npx @evalgate/sdk understand``npx skills add evalgate/skills --list` ## Release identity and maturity A source implementation, route, or test is not a general-availability claim. The platform is Beta; Agent Skills and MCP are Experimental. ### Application 3.18.2 ### npm TypeScript SDK 3.15.0 ### Python source / PyPI 3.18.2 / 3.15.0 ### OpenAPI contract 18.21.0 [Review the maintained feature-status inventory](https://www.evalgate.com/docs/platform/feature-status) ## Machine-readable agent references - [Documentation](https://www.evalgate.com/docs/) - [API reference](https://www.evalgate.com/docs/api/overview) - [OpenAPI](https://www.evalgate.com/openapi.json) - [Developer resources](https://www.evalgate.com/developers) - [MCP server](https://www.evalgate.com/docs/platform/mcp-integration) - [llms.txt](https://www.evalgate.com/llms.txt) - [Structured agent API view (JSON)](https://www.evalgate.com/?mode=agent) - [Public Skills repository](https://github.com/evalgate/skills) - [Public repository agent rules](https://github.com/evalgate/skills/blob/main/AGENTS.md) - [Public repository Agent Plugins manifest](https://github.com/evalgate/skills/blob/main/plugin.json) - [Public repository MCP configuration](https://github.com/evalgate/skills/blob/main/mcp.json) - [Runnable assertion test mode: GET descriptor, POST synthetic text](https://www.evalgate.com/api/demo/custom-eval) — no account, customer-state access, saved evaluation result, provider call or evaluation charge; normal access/API-usage metadata and quota counters apply. - [Agent Plugin manifest](https://www.evalgate.com/plugin.json) - [Agent Plugin MCP configuration](https://www.evalgate.com/mcp.json) - [Canonical Skill decision schema](https://www.evalgate.com/.well-known/agent-skills/evaluate-ai-change/assets/decision-contract.schema.json) - [Legacy Agent Plugin compatibility index](https://skills.sh/evalgate/agent-plugin) — external compatibility index; re-indexing to evalgate/skills is pending - MCP: Provides public read-only documentation and authorized, role/scope-filtered evaluation evidence and authoring tools.