Frigg is the open-source, serverless framework where an agent scaffolds a real, tested integration in an afternoon. Every integration it builds is also an MCP tool your own agents can call.
The connectors the community has already charted. Drop one in with a single frigg install.
Sometimes moving data is the whole job. Sometimes it's none of it. An integration might be an endpoint you expose, a queue you drain, a bit of provider-specific code, or a screen your user actually sees. Frigg treats those as first-class primitives, so you (and your agents) can reach for whatever the job needs.
Give it a path and you have a route. No side project, no separate service to stand up.
A quick annotation, and the background worker plus the infrastructure behind it come with it.
Attio, HubSpot, Zapier, Zendesk, Salesforce, and the rest. Frigg runs happily as the backend behind them, and the modules ship with helpers so your platform-specific code is built right.
Describe the in-app experience in a spec instead of hand-rolling it. Fenestra is the piece we're adding for integration UI, and it gets easier with every extension we ship.
Fenestra is an emerging spec we're introducing for in-app integration UI. It sits alongside the standards, it doesn't try to replace any of them.
Frigg has always been about not rebuilding the wheel for every integration. With an agent doing the wiring, the work moves to the parts that need judgment: what to connect, what to skip, and what the integration should actually do. The same structure that lets an agent build an integration also lets your agents use it.
Not boilerplate you copy. Run frigg init, then the agent harness grounds the model in your app's own conventions, queries the capabilities that already exist, scaffolds from templates, and validates its plan before it writes a line. You finish the last mile (mapping and business logic) instead of starting from a blank file.
Frigg deploys to your cloud, not a vendor's. It runs primarily on AWS today, and adopters have shipped it to GCP, Azure, and their own hardware via containers. Serverless means it costs next to nothing at rest and scales from three records a day to millions an hour on the same code. Customer data never transits someone else's servers, and there's no per-connection tax to renegotiate later.
Frigg describes what each integration does as typed capabilities like crm.contact.sync and notifications.slack.send. Flip the mcp-tool surface on and your agents call those integrations directly, right alongside raw MCP tools that are plain API requests. One capability graph, exposed to humans over GET /api/capabilities and to agents as tools.
Auth flows, pagination quirks, webhook signatures, the rate limit that isn't in the docs. Every API the community charts becomes a module anyone can install, so you start from what other people already worked out instead of re-reading the same reference page a hundred developers read before you.
Frigg is infrastructure-as-code. The app definition self-scaffolds the resources it needs (queues, functions, VPC-private networking, field-level encryption with your own KMS keys, OAuth2 wiring) as the integrations call for them. You describe what you want to integrate, and the cloud footprint is generated from that, so you're never hand-writing the same infra decisions twice.
The original vision was "spin one up in minutes, push to production in a day." With an agent doing the wiring, a first integration realistically lands in a few hours. Here's the actual path.
frigg init lays down the serverless app structure: handlers, infra-as-code, and the definition your integrations plug into.
frigg install hubspot pulls a charted API module, with auth, endpoints, and webhooks already mapped by the community.
The agent queries capabilities, copies a template, scaffolds tests, and validates its plan. You review and fill in the business logic.
frigg deploy stands it up on your own AWS account, VPC-private and encrypted, scaling on demand.
No magic, just a considered stack. Where you see an orange or, that's a real choice Frigg hands you instead of deciding for you.
It runs on infrastructure you own and stays out of your way about what you build on it. The use case is yours to pick. Here's where adopters have actually taken it.
Native, in-product integrations your end users authorize themselves. The CRM sync, the storage connector, the webhook that keeps two SaaS tools in step.
The plumbing behind your own operations. Moving records between the tools your team already runs, on a schedule or a trigger, without a per-task automation bill.
Frigg runs the same way for a weekend home-lab project as it does in production. Self and home automation, personal dashboards, whatever you feel like wiring together.
Frigg is a framework you add to a Node backend, not a platform you migrate onto. Try it on one integration this week.
One command to scaffold your first app:
The framework is open source. Use it, fork it, ship it, forever. Left Hook (the team that builds Frigg) also offers a commercial license for teams who want more than the community can give.
Join the Frigg Slack Connect channel for help, war stories, and early looks at the agent tooling. Or jump straight into GitHub Discussions.
