Connect your agent.

Agent Cert is platform-agnostic. Whatever framework your agent runs on, you can certify decisions with one API call or one MCP tool invocation. Pick your platform below for a walkthrough.

Developer CLIs (MCP-native)

• Claude Code — Anthropic's agentic CLI. claude mcp add or .mcp.json.
• OpenAI Codex CLI — OpenAI's terminal coding agent. codex mcp add or ~/.codex/config.toml.

Agent platforms

• OpenClaw — Native MCP client; alternative Skill (SKILL.md) for air-gapped setups.
• AWS Bedrock — Agent action group calling the REST API.
• Microsoft Copilot Studio — MCP server connector; native integration in Copilot Studio.

Frameworks & custom

• LangChain — Custom tool / callback hook.
• CrewAI — Tool wrapping the REST API.
• Custom agents — Any HTTP-capable runtime.

By transport

MCP (Model Context Protocol)

If your agent supports MCP, this is the fastest path — no custom code needed. Use it from Copilot Studio, Claude Desktop, or any MCP-compatible client. See MCP Server reference.

REST API

Single endpoint POST /v1/receipts. One header (X-API-Key), JSON body. Works from any language, any runtime. See REST API reference.

Choosing a path

If your agent runtime already speaks MCP: use MCP. It's less code, less to maintain, and tool discovery is automatic.

Otherwise, call the REST API directly. It's a single HTTP POST per decision — easy to drop into any callback / hook / middleware.