Guide

What is Oi MCP?

You can access your Agent Contexts through the Oi MCP (Model Context Protocol) server. This is the hosted surface for Oi. It gives MCP-capable clients a secure way to access Oi Agent Contexts.

If you are not familiar with this concept, it is a way to connect what you have setup within Oi; your agent contexts, workflows and so on - to any of your AI clients; Cursor, Claude, ChatGPT etc. that support MCP.

Core details

Server URL
https://api.oioioi.ai/mcp
Transport
Streamable HTTP over the hosted Oi endpoint.
Auth
OAuth or bearer token using an exported Oi organization API key.

Available Tools

These are the main MCP tools Oi exposes and what each one does. Some are tools you will use directly, and some support the experience behind the scenes.

MCP Tool: oi_use_agent_context

Use Agent Context

The main tool for doing real work with Oi. Use it when you want Oi to help with a task, either with a named Agent Context or by routing the task for you.

oi designer review this onboarding flow
use pv route this launch plan through the right Agent Contexts
use Oi product-manager turn this feature idea into a ticket
oi designer+growth-lead review this pricing page and suggest improvements
use pv designer@2 review these designs against the updated style guide

MCP Tool: oi_list_agent_contexts

List Agent Contexts

Shows the Agent Contexts your team can use. This is a good starting point when you want to see what is available before choosing one.

oi list the available Agent Contexts
use pv list agent contexts
use Oi list the available Agent Contexts

MCP Tool: oi_get_agent_context

Get Agent Context

Returns the compiled Agent Context itself. This is more useful when a client or workflow needs the reusable context directly instead of asking Oi to handle a task for you.

MCP Tool: oi_start_prompt_session

Start Prompt Session

Keeps one Agent Context active for the rest of the thread so follow-up requests stay in that mode.

Use Oi software-engineer in this thread for follow-up implementation work

MCP Tool: oi_recommend_agent_context

Recommend Agent Context

Suggests which Agent Context should help with a task when you want guidance before choosing one.

oi recommend the best Agent Context for reviewing this pricing page

MCP Tool: oi_save_feedback_to_draft

Save Feedback To Draft

Saves confirmed feedback into a private Agent Context draft so the context can improve over time.

MCP Tool: oi_confirm_guardrail

Confirm Guardrail

Handles guardrail confirmation when Oi needs an explicit user approval step before continuing.

MCP Tool: oi_report_agent_context_usage

Report Agent Context Usage

Reports usage metadata back to Oi after a client uses an Agent Context.