Developer Tools
HappyRobot is a fully programmable platform. Everything you can do in the UI, your engineering team can do through code: a documented REST API, SDKs, an MCP server, and workflows you can review in a pull request.
Book a demo- messageAgent00:02
- messageCaller00:19
- session_endedVisit booked00:47
Programmatic access
Build, trigger and manage every part of the platform from your own code, across more than twenty resource areas.
- REST APIWorkflows, versions, runs, sessions, phone numbers, integrations, knowledge bases and evals, with interactive OpenAPI docs.
- TypeScript SDKAn npm package for Node and TypeScript backends, with automatic retry and exponential backoff built in.
- MCP serverTwenty-plus platform tools exposed to any MCP-compatible environment, with domain context injected automatically.
Embed in your own product
Put a voice or text agent on any web surface, or bring your own telephony, without building infrastructure around it.
- Web SDKBrowser-based voice over WebRTC, a user clicks and speaks, no phone number required.
- Embedded chatA full text agent on any web property from one script tag, with file uploads of any type.
- Your own telephonyPhone numbers and SIP trunks are managed through the API, alongside direct CCaaS integrations.
Hi, I can help with your boiler. What’s happening?
No hot water since this morning.
boiler-photo.jpgWorkflows as code
Agent definitions are code, so they move through the same review and release process as the rest of your stack.
- Defined as codePrompts, nodes, tools and logic expressed as versioned code rather than clicks on a canvas.
- Reviewed in pull requestsEngineers build in their own environment and review changes the way they review everything else.
- Promoted through source controlReleases reach production through your existing branching and approval process.
- Workflow tests12 of 12 passed
- Staging run20 test calls
- ReviewApproved
Context as a data source
Query and write the same data layer your agents use during execution: live, not a nightly copy.
- MCP for ContextRead and write the tables your agents use, from any MCP-compatible development environment.
- Two-way REST APIA public API and managed gateway connect Context to your data lake, BI and analytics stack.
- Live, not exportedYou’re querying the layer written to during execution, so nothing waits on a batch job.
Thousands of workflows addressable visually or programmatically
Everything is addressable in code as well as on the canvas, so an engineering team can manage thousands of workflows the way they manage any other large system, without forcing the people closest to the work to become engineers.
Enterprise SuperintelligenceFAQs
Yes. The canvas is one way to author a workflow, not the only way. The same definition can live in your own repository and reach production through the process your team already uses for code.
That’s the intent of parity between the UI and the API: the same workflow, addressable either way. In practice an ops lead configures and iterates on the canvas while engineering handles the parts that want version control and review, and both are operating on the same definition rather than two copies that drift.
It exposes platform management tools like creating workflows, managing integrations, running evals, and configuring agents, so an assistant can do what a person with those permissions could do. The usual controls still apply: role-based access governs what the credential can reach, and promotion to production runs through an approval step rather than executing on request.
Yes. Phone numbers and SIP trunks are resource areas in the REST API, so provisioning and routing can be handled from your own systems rather than configured by hand. There are also direct integrations with several CCaaS providers where you’re bringing existing contact-center infrastructure.
Yes. Embedded chat supports uploads of any type like documents, images, audio and video, and the agent can read, transcribe and act on what it’s given rather than just acknowledging the attachment. It’s built on the same Web SDK as browser voice, so a page can offer both.
