Integrations
Connect the tools your work already lives in
One catalogue, 129 connectors across 16 categories — from your tracker and repo to your cloud, docs and models. Connect them in minutes, then bring their context and capabilities into your runs.
The catalogue
Find your stack
Start with the tools teams reach for first, or open the full catalogue. Adding a new one is a line of configuration — not a release.
Every connector is connectable today. Deeper capabilities —Memoryingestion andActions— are rolling out; Plane, GitHub and Slack lead.
GitHubDev & CI/CD
SlackCommunication
LinearProject management
NotionProductivity & docs
SalesforceCRM & sales
StripePayments
ShopifyE-commerce
Google DriveProductivity & docs
FigmaDesign & media
PostgreSQLDatabases
Anthropic (Claude)AI models
SentryDev & CI/CD
How it works
Connect, remember, act
An integration isn't a logo on a page. In TaskForce a tool earns its place in three steps — and each one is explicit about where it stands.
Connect
Connector availability
A connector can be configured and connected. OAuth is available for supported providers; the rest use encrypted credentials you provide.
Remember
Ingestion into Memory
A connected tool's data flows into the Brain OS decision graph, so runs can start from the context your team has already built. Live for Plane today, rolling out per tool.
Act
Action via MCP
Agents can use the tools you connect. Anything not explicitly read-only is treated as a write — the agent proposes it, and it runs only after a human approves it.
Illustrative — one shared memory, 129 connectors across 16 categories feeding every run.
Integrations brings in the context.
Memory preserves the reasoning.
Learn moreOrchestration decides what happens next.
Learn moreWe mark integrations Beta on purpose: connecting is here today, deep sync and MCP actions are rolling out. Follow it on the public roadmap.
Security & permissions
Your credentials, your scopes, your control
Connecting real systems to AI raises one question first: what can an agent actually do with that access? The short version is here — the full one lives in the Trust Center.
OAuth where supported
GitHub and Slack connect over OAuth; the rest use tokens or keys you provide.
Encrypted at rest
Every credential is stored encrypted, never in the codebase.
Bounded by capability
Agents can only invoke what a connector exposes — and only what the credentials and scopes you configure permit.
Out of the prompt
Connector secrets are used server-side to reach the tool; they're never placed in a model prompt.
Your tools become context and capabilities for every run
Connect your stack. Give every run the context it needs. Let agents act through the tools you already use — with humans still in control.