Skip to content
Canvas LabsCanvas Connectors

First working example

Goal: read a Linear issue from indexed knowledge and prepare a Slack-ready summary.

  1. Connect Linear as an indexed connector.
  2. Wait for the sync to complete.
  3. Connect Slack as an app-action provider if you want the agent to post live. Otherwise, ask for a draft.
  4. Use sketch_search with source: "linear" and kind: "task".
  5. Use sketch_get_file_content on the returned sketchId.
  6. Summarize the issue.
  7. Post to Slack through the live app-integration path, or return the draft to the caller.

The flow, in one line:

Read Linear issue -> summarize -> post Slack message

Ship it four ways

Section titled “Ship it four ways”

For a public cookbook, provide four versions — all in the SDK guides:

  • Claude Agent SDK — use Connectors as an HTTP MCP server.
  • OpenAI SDK — use the Responses API with an MCP tool.
  • LangChain — use @langchain/mcp-adapters.
  • Raw HTTP — call MCP JSON-RPC directly.