Self-hosted agent runtime. Tesserine runs autonomous AI agents on your own hardware — managing container lifecycle, enforcing cognitive methodology, and giving each agent an ephemeral workspace with exactly what it needs and nothing more.
| Component | Role |
|---|---|
| agentd | Daemon that runs agent sessions in ephemeral Podman containers. Owns identity, credentials, scheduling, and lifecycle. |
| runa | Cognitive runtime inside the container. Loads a methodology, validates artifacts, enforces the dependency graph. |
| groundwork | Open-source contributor methodology. The first methodology plugin for runa. |
| base | Reference container image. Wolfi-based, ships with the agentd contract and agent runtime tooling. |
| commons | Shared principles and design decisions across the ecosystem. |
Tesserine is a strongly typed cognitive state machine. Methodologies declare the topology — what artifacts exist, what protocols produce them, what conditions must hold before and after each transition. The runtime enforces fidelity: artifacts are validated against schemas, protocols fire only when their dependencies are satisfied, and the agent receives its work through typed interfaces that accept only well-formed output. The value comes from the guarantee that any declared cognitive process executes faithfully, regardless of its shape.
The operator declares what — which agent, which methodology, which project. The runtime owns how — container isolation, pipeline enforcement, artifact validation. Each layer is independently useful and independently replaceable: agentd manages the infrastructure, runa enforces the cognitive pipeline, and methodology plugins define how agents think and work.
v0.1.0 released. All ecosystem components tagged and under active testing.
From tessera (mosaic tile — composition from modular pieces), tesseract (higher-dimensional structure), and Madeleine L'Engle's verb to tesser (folding through dimensions). The adjective form says this is about how you do things. What you do is wide open.