Skip to main content

​
Prerequisites

  • A Correl8 account (app.correl8.ai)
  • An AI agent that supports MCP (Streamable HTTP) — or use one of the Examples as your agent

​
1. Create a project and API key

1

Sign up and create a workspace

Register at app.correl8.ai and complete onboarding to create your workspace and first project.
2

Generate an API key

Open Settings, create an API key, and copy it immediately — it is shown only once.Send it as Authorization: ApiKey c8.xxx.

​
2. Connect MCP

Your project MCP URL and project ID are in Settings → Details:
Configure the URL and Authorization: ApiKey c8.xxx header in your MCP client.

​
3. Add the Correl8 system prompt

Copy the short prompt from Agent system prompt into your agent’s system prompt. Your agent calls get_correl8_guidelines(user_id=…) once per conversation for posting rules. Pass stable user_id, interaction_id, and user_sentiment on each post_observations call.

​
4. Trigger a test observation

Chat with your agent and send a message that should produce a signal, for example:
The send button didn’t work when I pressed Enter.
The agent should call post_observations with user_sentiment and at least one observation (title and description).

​
5. Verify in Correl8

Open Observations in the app. New rows appear shortly after posting. Issues populate after server-side digest runs (usually within a few minutes). See Concepts for terminology, MCP for the full tool contract, and Examples for runnable integrations with Pydantic AI, LangChain, OpenAI Agents SDK, and REST.