Oi

Oi hosts a Streamable HTTP MCP server at api.oioioi.ai/mcp. Any MCP-compatible client — including ChatGPT, Claude, Codex, and Cursor — authenticates via OAuth or a bearer token using your organisation API key, then gets immediate access to all your team's Contexts, Guardrails, and Workflows without any per-client reconfiguration.

A Context is a reusable role-based prompt or operating instruction (e.g., 'Product Manager', 'Design Reviewer'). A Guardrail is a portable policy layer — rules like 'ask before publishing' or 'never leak sensitive data' — that attaches to any Context. A Workflow is an ordered sequence of Contexts with defined handoff steps, allowing multi-stage tasks like a full launch review to run end-to-end.

Yes. Any team can publish their private Contexts, Guardrails, Connections, and Workflows to the public Oi Marketplace. Other organisations can then discover, install, and fork those assets into their own private library to customise before deploying — enabling community-driven knowledge sharing across the platform.

Oi supports third-party integrations including GitHub and Linear as connectable data sources, allowing Contexts to be enriched with repository metadata, issue data, and product information. The number of active Connections depends on your plan (0 on Free, 1 on Pro, 3 on Team, 5 on Business).

Oi embeds usage and spend analytics directly into the platform, tracking agent message consumption, Context invocation counts, and operational logs per workspace. Analytics dashboards are available from the Team plan upward, giving administrators visibility into which Contexts are being used most and where token budgets are being spent.