Skip to content
Canvas LabsCanvas Connectors

Cookbook

The Canvas Connectors Cookbook is a public repo of runnable examples. Each recipe reads real context through the Connectors MCP server, then writes to a destination app (usually a Slack message), so you can see a full read → reason → act loop end to end.

The starter recipe reads a Linear issue, summarizes it, and posts the summary to Slack. It ships in four flavors so you can start from the stack you already use:

  • Claude Agent SDK — Connectors as an HTTP MCP server
  • OpenAI SDK — the Responses API with an MCP tool
  • LangChain — @langchain/mcp-adapters
  • Raw HTTP — MCP JSON-RPC directly, no agent framework

Recipes

Section titled “Recipes”
Recipe Command What it shows
Linear issue → Slack pnpm openai Remote MCP with sketch_search and sketch_get_file_content
Project brief → Slack pnpm project-brief Entity lookup with sketch_search_entities and sketch_get_entity_context
Meeting recap → Slack pnpm meeting-recap Recency search with kind: "meeting" and full transcript reads
PRD → Linear drafts pnpm prd-to-linear Document search plus safe-by-default Linear draft creation
ClickUp reminders → Slack pnpm clickup-reminders Source and task-kind filtering for operational digests
Linear release-risk digest pnpm linear-risk-digest Multi-search risk synthesis across Linear issues and projects

The starter recipe (Linear → Slack) is available in all four flavors under examples/; the additional recipes ship in OpenAI SDK and raw HTTP flavors.

The Connectors MCP server provides the indexed source context. The examples call destination app APIs directly for writes (a Slack post, an optional Linear issue), so the recipes work with the read-oriented public MCP surface available today.

Prerequisites

Section titled “Prerequisites”
  • Node.js 20+ and pnpm
  • A Canvas Connectors MCP URL and API key (see Getting started)
  • A synced Linear connector with at least one indexed issue
  • A Slack bot token with chat:write and a channel ID the bot is in
  • An Anthropic API key (Claude Agent SDK example) or OpenAI API key (OpenAI SDK, LangChain, raw HTTP)

Setup

Section titled “Setup”
Terminal window
git clone https://github.com/canvasxai/canvas-connectors-cookbook
cd canvas-connectors-cookbook
pnpm install
cp .env.example .env

Set your Connectors and destination credentials in .env:

Terminal window
CONNECTORS_MCP_URL=https://your-connectors-host/mcp
CONNECTORS_API_KEY=skp_...


LINEAR_ISSUE_QUERY=ENG-123
SLACK_BOT_TOKEN=xoxb-...
SLACK_CHANNEL_ID=C0123456789


ANTHROPIC_API_KEY=...        # Claude Agent SDK example
OPENAI_API_KEY=...           # OpenAI SDK, LangChain, raw HTTP
OPENAI_MODEL=...             # any model enabled for the OpenAI Responses API

LINEAR_ISSUE_QUERY can be an issue key, a title fragment, a project name, or any phrase that should find the issue. See .env.example for per-recipe variables.

Then run a recipe, for example:

Terminal window
pnpm openai          # Linear issue → Slack, OpenAI SDK

More in the repo

Section titled “More in the repo”

The cookbook also ships longer-form guides alongside the code: a Developer Guide, How Connectors Works, Security and Deployment, and SDK Notes.