pinter-agents) that allows external development environments and AI agents (such as Cursor, Claude Desktop, Zed, or custom MCP clients) to manage and interact with PINTER resources directly.
1. Connection & Transport
The server supports Streamable HTTP (recommended for production/remote IDEs), Server-Sent Events (SSE), and stdio (local CLI).Server Endpoint
2. Setup in External Clients
- Cursor
- Claude Desktop
- Local Process (stdio)
Add the server to your Cursor settings (Settings > Features > MCP Servers or in
.cursor/mcp.json):3. Available Tools Reference (29 Tools)
Thepinter-agents MCP server provides tools organized into 9 functional domains:
1. Agent Management
list_agents(search, status, limit, offset): Search and list agents visible to the authenticated user.get_agent(agent_id): Retrieve the full configuration, prompt instructions, and tool bindings of an agent.create_agent(name, model_id, description, identity, mission, methodology, boundaries, examples, tools, ...): Create a new agent owned by the caller.update_agent(agent_id, ...): Modify configuration fields of an existing agent.
2. Model Catalog
list_models(provider, active_only): List available and active LLM models (OpenAI, Anthropic, Gemini, local models).get_model_capabilities(provider, model_id): Inspect whether a specific model supports native thinking/reasoning.
3. Tool Catalog
list_tools(search, tool_type, limit, offset): Discover built-in and integration tools registered in the platform.get_tool(tool_id): Get full schema and connection settings for a tool.
4. MCP Server Registry
list_mcp_servers(search, transport, category, active_only, limit, offset): List approved external MCP servers.get_mcp_server(server_id): Inspect configuration and metadata of an MCP server.test_mcp_server_connection(server_id): Perform a live handshake test to verify connectivity and tool discovery.
5. Asset Management (Rules & Skills)
list_assets(asset_type, search, folder_path, limit, offset): List rules, skills, or knowledge assets.get_asset(asset_id): Retrieve complete content markdown and metadata of an asset.create_asset(type, name, content_markdown, description, visibility, ...): Create a new versioned skill or rule.update_asset(asset_id, name, content_markdown, visibility, ...): Publish a new version of an existing asset.
6. Asset & Document Bindings
list_agent_asset_bindings(agent_id): List all rules, skills, and knowledge documents attached to an agent.replace_agent_asset_bindings(agent_id, bindings): Attach or re-order assets for an agent, configuring usage modes (always_on,on_demand,retrieval).
7. System Announcements (Platform Architect / L1 Only)
list_announcements(search, severity, status, limit, offset): List broadcast announcements.create_announcement(title, content_text, severity, starts_at, ends_at): Create a draft announcement.update_announcement(announcement_id, title, content_text, severity, ...): Update announcement content and schedule.delete_announcement(announcement_id): Soft-delete an announcement.
8. Quotas & Budget Requests
get_my_quota(): Inspect your active token limits, monthly budget, and remaining balance.create_budget_request(requested_amount, reason): Submit a budget increase request for review.list_my_budget_requests(status, limit, offset): View status of your past budget top-up requests.review_budget_request(request_id, status, approved_amount, admin_note): Approve or reject a budget request (requiresquota:managepermission).
9. Agent Evaluations
list_eval_sets(search, agent_id, limit, offset): List golden dataset evaluation suites.get_eval_set(eval_set_id): View test cases and scoring rubrics.create_eval_set(name, description, agent_id, eval_cases, is_public, criteria): Create a new evaluation benchmark.run_agent_eval(eval_set_id, agent_id): Execute evaluation benchmarks against the agent and return pass rate and scores.