Agents and MCP
Use AI agents and the Model Context Protocol with OBP.
AI Agents and the Model Context Protocol
The Open Bank Project supports integration with AI agents through the Model Context Protocol (MCP). MCP is an open standard that allows AI assistants and agents to discover and call APIs in a structured way.
What is MCP?
The Model Context Protocol provides a standardised way for AI models to interact with external tools and data sources. Instead of hard-coding API knowledge into an AI model, MCP lets the model discover available tools at runtime and call them with the correct parameters.
Think of it as a universal adapter between AI agents and APIs. The agent asks "what can I do?" and the MCP server responds with a list of available tools, their parameters, and descriptions.
OBP MCP Server
The OBP MCP server exposes the OBP API as a set of tools that any MCP-compatible AI agent can use. It provides:
- Endpoint discovery — list all available API endpoints, filtered by tag
- Schema inspection — get the full request/response schema for any endpoint
- API calls — make authenticated calls to the OBP API
- Glossary access — look up OBP terminology and concepts
Available MCP Tools
| Tool | Description |
|---|---|
list_all_endpoint_tags | List all API tags (categories) to discover available endpoint groups |
list_endpoints_by_tag | List endpoints within a specific tag/category |
get_endpoint_schema | Get the full schema (parameters, request body, response) for an endpoint |
call_obp_api | Make an authenticated API call to any OBP endpoint |
list_glossary_terms | List all glossary terms and their definitions |
get_glossary_term | Get the full definition of a specific glossary term |
Connecting a client
The OBP MCP server is a remote HTTP server — you connect to it by URL, there is nothing to install. This instance's server is at:
https://mcp.dcr.ogcr.tesobe.comAdd it to an MCP client such as Claude Code:
claude mcp add --transport http obp https://mcp.dcr.ogcr.tesobe.com
# then sign in (opens your browser):
claude mcp login obpOr add it directly to the client's configuration file (for example .mcp.json for Claude Code / Claude Desktop):
{
"mcpServers": {
"obp": {
"type": "http",
"url": "https://mcp.dcr.ogcr.tesobe.com"
}
}
}Authentication uses OAuth with your OBP account. On first connect the client opens a browser sign-in, then stores and refreshes the access token for you — there is no token to paste into the configuration. The server then calls the OBP API as you, limited to what your account is permitted to do.
Once connected, the AI agent can discover and call OBP API endpoints autonomously. For example, you could ask Claude: “List the banks available on this OBP instance” and it will use the MCP tools to make the API call and return the results.
Using with Other AI Agents
Any AI agent or framework that supports MCP can connect to the OBP MCP server. This includes:
- Claude Desktop — add the MCP server in settings
- Claude Code — configure via
.mcp.jsonin your project - Custom agents — use the MCP SDK to connect programmatically
Building Your Own Agent
You can build custom AI agents that interact with the OBP API by:
- Connecting to the OBP MCP server by URL and signing in (see above)
- Using an MCP client library to connect your agent to the server
- Letting your agent discover available tools and make API calls
This enables powerful use cases like automated compliance checking, transaction monitoring, customer onboarding workflows, and natural language banking interfaces.
Resources
- Model Context Protocol specification
- OBP GitHub organisation
- Opey — the built-in OBP AI assistant
- What Opey Can Do — consent-based authorisation for agents, explained
- OAuth2 / OIDC — how sign-in to the MCP server works
- API Explorer — browse all available endpoints