Connect to everything.
APIs, CLIs, MCP servers, or a headless browser — four ways to connect your apps and agents to the tools you already use. And if a tool has no API, your agent can just use it like a person would.
RedisMongoDBShopifyXeroFour ways to connect
Not just API wrappers. Real integration methods that cover every scenario — from standard REST calls to agents that drive a web browser.
REST APIs & webhooks
Connect to any service with an API. Your apps call external endpoints directly — Stripe for payments, Twilio for SMS, HubSpot for CRM. Standard HTTP, no middleware.
- Direct HTTP calls from your app
- Webhook receivers for inbound events
- Environment-scoped API keys
- Any REST or GraphQL endpoint
CLI tools
Every workspace has a full terminal. Install and run any CLI — database clients, cloud tools, build systems, linters. If it runs in a terminal, it runs on Shaep.
- Full terminal in every workspace
- Install any package or binary
- Cron jobs and scheduled scripts
- Pipeline steps with custom CLIs
MCP servers
Model Context Protocol lets your AI assistants talk to external tools directly. Connect Claude, GPT, or any MCP-compatible model to your databases, APIs, and internal systems.
- AI assistants call your tools
- Connect to databases and APIs
- Custom MCP servers for internal systems
- Bring any MCP-compatible model
Headless browser
No API? No problem. Agents interact with any web application through a headless browser — fill forms, scrape data, trigger workflows. If a person can use it, an agent can too.
- Access apps that have no API
- Fill forms, click buttons, extract data
- Automated testing of deployed apps
- Agents interact with any web UI
No API? No problem.
Some tools don't have APIs. Legacy systems, internal portals, government platforms. Your agents interact with them through a headless browser — navigating pages, filling forms, extracting data. If a person can use it, an agent can automate it.
- Works with any web application
- Full audit trail of every action
- Scheduled or on-demand
Agents that act
Integrations aren't just for your apps. Autonomous agents use all four methods — with their own identity, scoped permissions, and a full audit trail.
Agent gets a task
"Every morning, check the fleet portal for new deliveries and create routes in the planner."
Agent connects
Opens the portal via headless browser, calls the planner API, updates the database. Uses whichever method each system requires.
Everything is audited
Every API call, browser action, database write — logged. Scoped permissions ensure the agent only touches what it's allowed to.
Frequently asked questions
Do I need to write code to integrate? +
Not necessarily. The AI assistant can set up integrations based on your description. For APIs, it generates the connection code. For MCP, it configures the server. For the headless browser, it writes the automation script.
What about authentication? +
API keys and secrets are stored encrypted per app, scoped to the environment (dev/prod). They're never exposed in logs, diffs, or the UI. Rotate them without redeploying.
Can agents run integrations autonomously? +
Yes. Agents can be deployed with scoped permissions and scheduled tasks. They can call APIs, run CLI commands, connect via MCP, or use the headless browser — all with a full audit trail.
What if the service I need doesn't have an API? +
That's exactly what the headless browser is for. Your agent opens the web application, navigates the UI, fills forms, and extracts data — just like a person would, but automated and auditable.