Tools & MCP is the connector surface for harness agents — built-in platform tools, the Exemplar MCP endpoint, and connector A2A agent cards so agents can act on real systems with credentials you control
Exemplar
How this harness capability fits the Exemplar platform—governed agent operations, not a standalone prompt playground.
Without a shared tool plane, every agent team hand-wires Jira, PagerDuty, Datadog, and cloud APIs—different auth, no shared audit, token-heavy glue.
Exemplar publishes connectors as unified MCP tools on one governed endpoint, so frameworks and IDE clients share the same catalog, policy, and audit trail.
A flat catalog of unified tools with one { action, params } call shape—platform tools plus connector *_tool actions on {agent_base}/mcp.
A2A specialist agents per connector—agent cards and JSON-RPC endpoints orchestrators can hand a whole turn to.
Enable connectors; unified tools appear on the org MCP endpoint for Cursor, Claude, and harness.tools() without per-vendor SDKs.
Import your own OpenAPI specs to publish custom endpoints as MCP tools behind the same governance.
Official documentation on docs.exemplar.dev for this capability.
Open developer guide (opens in a new tab)Contact sales
Harness Platform is scoped per deployment. Talk to us about this feature.
Related posts on exemplar.dev.
Model Context Protocol (MCP) is the open standard that lets AI assistants connect to your tools and data. How it works, how it differs from function calling and RAG, which tools support it, and what it means for production engineering.
The best AI agent control plane and harness tools for running agents in production — governing actions, managing token costs, orchestrating workflows, and keeping a full audit trail. Compared and ranked.
From code completion to production actions; Context Lake, catalog, governance, and DevX Assist/MCP for safe automation.