The open platform for tracked agent execution
A self-hosted platform for running isolated sandboxes, capturing execution history, and building modular developer workflows on top.
Issue Workflow
Capture issues
Pull issues from GitHub or Linear, and define the execution outline
Image Creation
Define your image
Drop in environment details, build inputs, and policies for your sandbox lifecycle.
Clone repo, install dependencies
Agent work
Execute your changes
Use your preferred harness, and make desired changes
Output
View Execution Summary, Open PR
View execution metrics, outcome summaries, or a polished trace snapshot.
Metrics
Platform overview
Tracked software work, built on two core primitives.
Sealant centers two core primitives: sandboxes and executions. Workflow modules build on top of that foundation.
- The Core
Two core primitives: fast, highly customizable sandboxes and tracked software runs (executions). Sandboxes provide the isolated environment. Executions provide the durable record of what ran, what changed, and how the run completed.
- Observability
Execution visibility is a core product surface. Sealant focuses first on state transitions, artifacts, diffs, and run summaries, with deeper tracing added over time.
- First-party modules
Sealant is designed to support first-party workflow modules, starting with Issue Workflows. These modules are intended to exercise the same execution, artifact, and policy surfaces that future extension points will build on.
- SDK & extensions
Sealant is being built with future extension seams in mind. Early extension points are expected around workflow hooks, reporters, runtime adapters, and first-party- style modules, with the public SDK formalized after those seams are proven in product use.
Open Source
Self-hosted by default.
Run Sealant inside your own boundary and build on an open platform for isolated, tracked software work.