Skip to main content

​
What is the Tools module?

The Tools module is Connie’s connectivity layer. It’s where every external connection and custom integration lives — and where your workflows and AI copilot reach when they need to do something outside the product.

Integrations

OAuth connections to standard SaaS — Slack, GitHub, HubSpot, Salesforce, and ~30 more. Connect once, use everywhere.

MCP servers

Custom or self-hosted Model Context Protocol servers. For anything the integration catalog doesn’t cover.

Webhooks

Inbound endpoints that trigger workflows when external services emit events.

​
Integrations

External SaaS connections — OAuth, token refresh, and credential storage — are managed for you. From the user’s perspective:
  1. Open Tools → Integrations, pick a provider
  2. Click Connect → standard OAuth popup with the provider
  3. The connection lands in your workspace and is immediately usable by workflows, the AI, and other modules
Connecting and disconnecting accounts happens through the in-app flow (it’s an interactive OAuth handshake, not a headless API call). What the API exposes is the catalog of available providers, so you can see what can be connected:

​
What’s in the catalog today

~32 providers ship with Connie out of the box. The most-used: Each connected account is per-workspace and shared across the team — once an admin connects a Slack workspace, every member’s workflows can post to it. Members can also create their own accounts where applicable (e.g., a personal Gmail), with admin-controlled access modes per tool: auto, requires_approval, or disabled.

​
MCP servers

MCP (Model Context Protocol) is the open standard for exposing tools and data to AI agents. Connie has full first-class support for MCP — you can attach any MCP-compliant server to your workspace and its tools become available to the AI brain alongside the built-in integrations.

​
When to pick MCP vs an integration

​
Two flavors of MCP support

Hosted MCP (OAuth)

Connect to MCP servers exposed by vendors via OAuth — same UX as a catalog integration. Tools are discovered automatically.

Direct / custom MCP

Point Connie at any HTTP MCP endpoint with custom auth — env vars, basic auth, or bearer token.

​
Per-tool enable / disable

Once a server is connected, every tool it exposes appears in the workspace tool list. Admins can toggle individual tools on or off — the AI will only consider the enabled ones.

​
Webhooks

Webhooks let external services push events into Connie — triggering workflows in real time. A subscription exposes an inbound endpoint that verifies the event and routes it through Connie’s trigger system:
Once an event arrives:
  1. Its signature is verified against the subscription’s secret
  2. The event is parsed and routed through Connie’s trigger system
  3. Any workflow subscribed to that event fires with the event payload
This is the most common way to wire “when X happens in Service Y, do Z in Connie” automations.

​
Common workflows

​
Connect Slack and post workflow alerts

  1. Tools → Integrations → Slack → connect
  2. Build a workflow with an on workflow failure trigger
  3. Add a Slack node, pick a channel, write your alert template
  4. Save — every failure now pings the channel

​
Add a custom MCP server for your internal API

  1. Tools → MCP → Add server → choose “Direct / custom”
  2. Paste your MCP endpoint URL and auth (env var or bearer token)
  3. Connie discovers the tools; toggle the ones you want exposed
  4. The AI copilot can now call your internal tools in any conversation

​
API reference

Integration providers

Browse the catalog of available providers.

MCP servers

Add and manage Model Context Protocol servers.

MCP tools

Discover and toggle individual tools from a connected MCP server.

Webhooks

Inbound webhook endpoints that trigger workflows.