The building blocks¶
Every agent in DAM is assembled from the same reusable pieces — compute, credentials, skills, memory, schedules, and more. These are the plumbing you'd otherwise rebuild for each new project, on one laptop at a time. DAM ships the pieces and you compose whatever your team needs, with no opinion baked in about what you build. Here's what you have to work with.
- Isolated compute + persistent workspace A long-running, isolated Linux sandbox with a
~/workdirectory that survives restarts and hibernation — the agent keeps working after you close your laptop. See Agents. - Network boundary & credential gateway The agent reaches only the services you allow, through a hardened gateway that holds your tokens and injects them on the wire — it never sees the raw credential. See Network & Approvals.
- Connections Wire in the accounts an agent acts through — GitHub, API keys, and more — without it ever seeing your tokens. See Connections.
- Skills Packaged instructions that teach an agent a capability — plan a feature, draft a PRD, file an issue — installed once and reused. See Skills.
- Schedules Run an agent unattended on a recurring timer — daily reviews, nightly audits, continuous monitoring — with no human in the loop. See Schedules.
- Channels Reach a running agent where the team already works — a Slack channel or DM, or a Telegram chat. See Slack and Telegram.
- Sessions Persistent chat and terminal conversations with an agent — the thread survives across restarts and hibernation. See Sessions.
- Files & memory Browse, edit, and upload files in the persistent workspace that holds the agent's configuration and memory across sessions. See File Manager.
- Artifacts Finished outputs an agent publishes to keep — a report, a dashboard, a generated page, a data file — that outlive any single run. See Artifacts.
- Knowledge bases An agent whose job is to hold a body of knowledge you chat with, answering with citations back to the source. See Building knowledge bases.
- Starter kits The blocks above, pre-assembled into a ready-made agent you create in one step and it sets itself up. See Starter kits.
The blocks are the point. A coding assistant that reviews pull requests, a research agent that runs experiments, a knowledge base your whole team can query — these are all the same building blocks combined differently. DAM gives you the pieces; what you build with them is yours.
Put them together and you get the things our team actually runs today — a personal brain you DM, an automated code reviewer, a live optimization loop. Several of those are offered as starter kits, so you can have one without assembling it yourself; the guides walk through building a few end to end.