One endpoint for your whole messaging stack
Any MCP-compatible client or agent can work with your Title brand in natural language. Here is the shape of the server.
Prefer plain HTTP? The same brand is reachable through the REST API. To plug your own AI agent into a text line, see Agent Connect.
Add the server to your client
Pick whichever your client supports: paste the remote URL, or drop a config block into your config file. Then sign in once.
https://titlebm.com/api/mcp139 built-in tools
Tools marked write need the write scope (or admin). The second pill is the tool's permission group: the person who connected the AI (or their admin) can switch a whole group off in Settings → AI connections, and the tool then answers FORBIDDEN naming the switch. Each connected integration adds its own tools to the same session, for example a HubSpot connection exposes hubspot.* tools alongside these.
Tool results are JSON. A failed call returns an error result with a code such as INVALID_PARAMS, NOT_FOUND, UNAUTHORIZED or RATE_LIMITED. send_message, trigger_flow and create_campaign send real messages to real people; the server asks the model to confirm with you first.
Browsable data
Read-only MCP resources let an agent page through data without a tool call. The URI templates accept query parameters.
11 prompts
Task prompts steer the model toward the right sequence of tools. Skill prompts are reference material the client loads when it needs it, so the full guide is never in context by default.
10 skills that teach any AI how to use Title
Short, generic guides written for an AI: when to use the skill, a quick reference of the tools and the permission group each needs, step-by-step recipes and common mistakes. Never specific to one kind of business.
Through MCP, nothing to install: a connected AI calls list_skills and get_skill, or reads title://skills/<slug> as a resource. Claude Code: unzip and copy the folders into .claude/skills/. Claude.ai: Settings → Capabilities → Skills → upload one folder at a time (/api/skills/download?skill=<slug> downloads a single skill). Any other AI: the files are plain markdown; paste the one you need into its instructions.
OAuth 2.1, one connection per person
The server is a standard OAuth 2.1 protected resource. An unauthenticated request gets a 401 with discovery metadata; the client registers itself and sends you to sign in. No keys to manage.
Same endpoint, Agent Connect key
Agent Connect lets a business plug its own AI agent into its text line. Send that connection's API key as the bearer token instead of signing in, and the server exposes only the six agent tools, scoped to that one connection.
OAuth sessions never see these tools, and an agent key never sees the tools above. The rules and error codes match the REST endpoints. Setup, events and guardrails are in the Agent Connect guide.
Per-organization limits
Going over a tool limit returns an error result with code RATE_LIMITED so your agent can back off. The HTTP limit answers 429 with a Retry-After header.
https://titlebm.com/api/mcp