Blueticked
Start free

MCP server

Blueticked ships a remote Model Context Protocol server, so a coding agent can call the API as tools instead of writing HTTP by hand. It is hosted — there is nothing to install, and the tool list can never be a version behind the API.

Connect it

Claude Code
claude mcp add --transport http --scope user blueticked https://app.blueticked.com/api/mcp
Codex
codex mcp add blueticked --url https://app.blueticked.com/api/mcp
codex mcp login blueticked

Claude Desktop and claude.ai (paid plans): open Settings → Connectors → Add custom connector and paste https://app.blueticked.com/api/mcp. Claude opens the Blueticked sign-in and the same approval screen as every other client.

Follow any setup path, sign in to Blueticked in the browser, and approve a Test connection while you build. You choose the organization, which businesses the agent may touch, an access profile, Test or Live, and an expiry. Your client stores and refreshes its OAuth tokens; no API key is copied into the project.

Knowledge built into the connection

  • Complete API resource — blueticked://docs/full contains authentication, scopes, endpoints, examples, errors, and webhook rules in one read.
  • Machine-readable contract — blueticked://openapi exposes the same generated OpenAPI 3.1 document used by the Developer Hub and SDK workflow.
  • Drop-in project instructions — AGENTS.md, CLAUDE.md, SKILL.md, and production integration recipes are available as MCP resources and public downloads.
  • Guided implementation prompts — built-in prompts cover SMS OTP, WhatsApp notifications, transactional email, signed webhooks, and production-readiness audits.

What your agent can do

Typed tools for the work integrations actually do — send a message, look up or upsert a contact, read a conversation, check delivery status, manage templates and webhooks — plus a discovery tool and an escape hatch that reaches anything else: list_endpoints returns the required fields and a working example body for whatever you are about to call, and call_api reaches any public endpoint. The live tool list ships with the connection itself and in llms-full.txt.

Platforms serving their own customers can drive the whole onboarding through the agent too: create a business, mint its scoped credential, and send its owner a WhatsApp connect link — no step needs a human in the Blueticked dashboard.

What it can and cannot reach

  • Delegated, never wider — the OAuth token only reaches the MCP endpoint. Blueticked exchanges it for the encrypted Agent key created by your approval, then the same API scopes, business restrictions and rate limits apply.
  • No second permission model — Observe, Prepare and Operate profiles resolve to fixed API scopes on the server. The delegated key is encrypted at rest and never returned to the client. If its API request would return 403, the MCP tool does too. Every tool also advertises its route and required scope in _meta, so an agent can tell what it may call before calling it.
  • Failures come back readable— an API error is returned as the tool's result, not swallowed into a protocol error, so the agent can read the code and correct itself.
  • Reachable only inside /api/v1 — the escape-hatch tool takes a path, and anything outside the public API — absolute URLs, traversal, internal routes — is refused before a request is made.

Build against the sandbox first

Approve a Test connection and the agent can exercise the whole integration — sends, webhooks, failures and inbound replies — without touching a provider or spending anything. See test mode and magic numbers.

Plain-text context for agents

If you would rather hand your agent the whole contract than connect a server, these are public and need no credentials:

  • llms.txt — the index: what Blueticked is, where everything lives.
  • llms-full.txt — the complete contract in one document: auth, conventions, every endpoint, scopes, error codes, webhooks and the known pitfalls. This is the one to fetch if you are writing code.
  • openapi.json — OpenAPI 3.1, the authoritative machine-readable schema.
  • agents.md — a dashboard map, for agents driving the UI rather than the API.