Skip to main content
AI agents can create projects, add keywords, run checks, and read rank history. They use the same authenticated operations as other API clients.

​
What an agent can do

  • Connect and test a SERP provider.
  • Add keywords to a project.
  • Run a rank check and read the result.
  • Read ranking history and stored traffic snapshots.
  • Review, mute, or acknowledge triggered alerts.
  • Export project data or prepare a migration.

​
Connect through MCP

bisibility exposes a Model Context Protocol (MCP) Streamable HTTP endpoint:
Use https://bisibility.com/api/mcp for the hosted production endpoint. A self-hosted client should use its own origin. Hosted MCP clients that implement the MCP OAuth flow discover bisibility’s authorization server automatically. The flow uses Authorization Code with PKCE, dynamic client registration, and access tokens bound to the /api/mcp resource. The user reviews read, write, and admin scopes before approval. Clients that manage bearer credentials outside the MCP OAuth flow can still use a personal access token or project API key. For example, connect Codex to a self-hosted instance with a personal access token:
Project API keys work too. Other MCP clients use different configuration fields. The MCP server card lists the endpoint and available tools:

​
Authenticate

Choose a personal access token when the agent acts as you across projects or performs account-level operations. Choose a project API key when the agent should stay isolated to one project. Send either credential in the HTTP Authorization header, not in a tool argument. For hosted interactive clients, prefer the OAuth flow so the client never asks the user to copy a long-lived API credential.
For REST clients and SDKs, set the API root and credential in the tool environment. The BISIBILITY_API_KEY variable accepts either credential:
See Authentication to choose and create a credential, select a project for PAT requests, and make a test request.

​
Example workflow

  1. Create or select a project.
  2. Add a keyword with its country and device.
  3. Start a rank check.
  4. Poll the check until it completes.
  5. Read the keyword’s ranking history.
MCP tools use the same permission model as REST API v1 and expose the same documented operations.

​
More agent tasks

  • Discover keywords that a domain already ranks for and annotate tracked terms.
  • Research related queries, suggestions, and keyword ideas, then hydrate their metrics.
  • Query live search-performance data or start an analytics sync.
  • Generate a weekly report of ranking gains, losses, and new entries.
  • Inspect, enable, or disable sitemap monitoring for a project.

​
Available tools and schemas

MCP server card

The MCP endpoint and tool inventory.

OpenAPI

The exact REST API schema.

Capabilities

Features exposed by the production API.

Agent skills

Task-focused instructions that agents can discover.

​
Advanced discovery

These endpoints help agents discover authentication, schemas, and supported features. They are optional for a basic MCP connection.

​
Optional DNS-AID records

HTTP discovery works without DNS-AID. Operators can publish DNS records when they also want domain-level AI discovery:
DNS-AID and DNSSEC are separate operator settings. Sign the zone and publish the matching DS record before claiming authenticated DNS discovery. Verify the records with a validating resolver.