Connect your AI assistant (MCP)
BindCheck ships a Model Context Protocol (MCP) server, so AI assistants that speak MCP — Claude, and a growing list of agent tools — can work with your checks directly: list recent renewal checks, read the findings on one, fetch the diff summary, or ask free-form questions. Same keys, same permissions, same limits as the REST API.
Connecting
The endpoint is https://app.bindcheck.com/api/mcp (streamable HTTP). Authenticate with an API key from Settings → API as a bearer token — API access is included in the Agency plan. In Claude Code: claude mcp add bindcheck https://app.bindcheck.com/api/mcp --transport http --header "Authorization: Bearer bindcheck_…". Most MCP clients accept the same URL + Authorization header pair.
What the assistant can do
The tools mirror the dashboard:
- list_checks — filter by status (processing, done, error), kind (renewal, quote), review-required or date.
- get_check — the full picture for one check: the diff summary and every finding with severity, category and remediation.
- get_policy — the extracted forms, fields, limits and dates for one uploaded policy.
- checks_summary — org-wide counts at a glance (open findings, checks awaiting review).
- ask_checks — free-form questions ("which renewals this month dropped an additional-insured endorsement?").
Permissions and limits
Read-only API keys expose only the read tools — an assistant with a read key can never modify anything. All MCP traffic counts against the same per-key burst limit and daily API budget as the REST API, and every action is attributed to the key in the activity log, exactly like REST calls.