generated: '2026-08-12'
method: searched
source: https://www.agentio.com/connector
status: published
server:
  name: Agentio Brand Connector
  transport: http
  url: https://mcp.agentio.com/mcp
  protocol: MCP (Model Context Protocol) over JSON-RPC 2.0
  hosting: 'vendor-hosted (remote); Google Cloud (Server header: Google Frontend)'
  docs: https://www.agentio.com/connector
  support: support@agentio.com
access:
  mode: read-only
  scoping: one brand per connection; the brand is chosen during the OAuth consent step
  stated_limits:
  - cannot change anything in the account
  - cannot reach another brand's data
  - cannot act on the user's behalf
  prerequisites:
  - an Agentio account (sign-in happens during connector setup)
  - an MCP client that supports custom connectors (docs name Claude and ChatGPT; any MCP client works
    the same way)
  - availability depends on the customer's Agentio plan, and a workspace admin may control which connectors
    can be added
  management: Settings -> Integrations in the Agentio app lists every connected AI app and when it connected;
    disconnecting takes effect immediately. Each app holds one connection.
authorization:
  scheme: OAuth 2.1 authorization code + PKCE
  challenge: 'anonymous requests return HTTP 401 with WWW-Authenticate: Bearer resource_metadata="https://mcp.agentio.com/.well-known/oauth-protected-resource/mcp"'
  protected_resource_metadata: well-known/agentio-oauth-protected-resource.json
  authorization_server_metadata: well-known/agentio-oauth-authorization-server.json
  scopes:
  - brand-connector:read
  profile: authentication/agentio-authentication.yml
tools:
  introspection: gated
  note: 'NOT ENUMERATED. An anonymous JSON-RPC tools/list POST to https://mcp.agentio.com/mcp returns
    HTTP 401 {"error": "invalid_token"}, and initialize returns 401 as well, so the live tool names and
    inputSchemas require an authenticated introspection with a brand-connector:read token. Agentio publishes
    no llms.txt and no tool reference page, so there is no published tool list to fall back on. No tool
    names are recorded here, because inventing them would fabricate the provider''s contract.'
  documented_capabilities:
  - read campaigns
  - read creator deals and deliverables
  - read YouTube ad performance
  - read conversion reporting
  - explain how Agentio works (metric definitions, how a workflow runs)
  documented_example_questions:
  - Which of my campaigns are running right now?
  - How did my YouTube ads perform last month, broken down by campaign?
  - What did we spend per conversion last quarter, and which sources did those conversions come from?
x-evidence:
  fetched: '2026-08-12'
  probes:
  - url: https://mcp.agentio.com/mcp
    method: POST tools/list
    http_status: 401
    body: '{"error": "invalid_token"}'
  - url: https://mcp.agentio.com/mcp
    method: POST initialize
    http_status: 401
  - url: https://mcp.agentio.com/.well-known/oauth-protected-resource/mcp
    http_status: 200
  - url: https://www.agentio.com/connector
    http_status: 200
deployment:
  mode: remote
  endpoint: https://mcp.agentio.com/mcp
  verified: probed
  probe: wall
  checked: '2026-08-12'
  source: catalog MCP census