- Category
- AI Agents
- Rank
- No. 3603Tools index
Previous survey · No. 3428 ·
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Coding
- Interfaces
- CLI · Web
- GitHub
- 72 stars
- Latest release
- v0.14.0-rc.2
- Date
About
engrams is a self-hosted orchestrator for AI coding agents. Each agent runs in its own Firecracker microVM built from a plain OCI image, which is snapshotted when the agent goes idle and restored on the next prompt. Transcripts and tool calls stream to a dashboard, Slack, or CLI, and sessions run on hosts in the user's own cloud.
At a glance
engrams runs coding agents in Firecracker microVMs on cloud infrastructure you control. It is worth evaluating when agents need isolated workspaces they can resume after going idle.
Good fit
Infrastructure teams that can operate the deployment, customize guest images and set network policy.
Consider something else when
Use a lighter setup when you only need occasional local agent sessions or cannot maintain the virtualization infrastructure.
Start here
Run a public test project, leave a file in the guest and let the session idle. Resume it and verify both the file and an outbound request that policy should block.
Cost and requirements
The self-hosted project documents cloud Kubernetes and Firecracker infrastructure. Budget for compute, snapshot storage and inference.
Limits to know
Network filtering and VM isolation depend on deployment configuration. Snapshots, logs and integrations can retain sensitive data, so review where those records go.
Sources checked 2026-11-05. Source-based assessment.
Intel on engrams
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.