osFoundry / Platform
Agent Sessions
See every long-running agent at a glance: who is working, who is waiting, who is done.
What it is
Agent Sessions is the live status view for long-running agents inside osFoundry. When you hand a multi-step job to an agent — a coding task, a research run, a data pull — it can work for minutes or hours. Instead of reopening each conversation to check on it, Agent Sessions gives you one running list of every active agent and its current state: who is working, who is waiting, and who is done.
Because osFoundry is a bring-your-own-key desktop workspace, your agents run against your own model keys — local, cloud, or self-hosted — with no reseller markup. Agent Sessions tracks them all in the same view regardless of which model backs each run. The "waiting" state is the part that earns its place: when an agent pauses for an approval on a sensitive action or needs an answer from you, it surfaces there instead of stalling silently, so you can jump in and unblock it. Open any session to follow its progress in detail, then get back to the overview.
This matters most when you are running more than one agent at a time. Long runs are easy to lose track of, and the failure mode is usually an agent quietly sitting on a question you never saw. A single at-a-glance board of working, waiting, and done states removes that blind spot: you can start several agents, walk away, and know exactly which one needs you the moment it does — without paying per-seat fees or shipping your data off your device to see it.
Key capabilities
- Live overview of every long-running agent in one place, with three clear states: working, waiting, done
- Waiting agents surface the instant they need an approval or your input, so a run never stalls unnoticed
- Tracks agents across local, cloud, and self-hosted models from one desktop view — your keys, no markup
- Run many agents concurrently and check on them without reopening each conversation
- Open any session to follow its progress, then return to the board and hand off the next task
Use cases
- Solo builder: starts three coding agents on separate branches and watches the board to see which one is blocked waiting for an approval
- Researcher: leaves several long research agents running overnight and, in the morning, sees at a glance which have finished and which are still working
- Ops lead: monitors a set of scheduled agents and jumps straight to the one stuck waiting on input instead of polling each run
Frequently asked questions
What does the "waiting" state actually mean?
An agent is waiting when it has paused because it needs something from you — approval for a sensitive action, an answer to a question, or a decision before it can continue. Agent Sessions puts those agents front and center so you can unblock them quickly rather than discovering later that a run sat idle.
Does this work with any model I use?
Yes. osFoundry is bring-your-own-key, so agents run against your own local, cloud, or self-hosted models, and Agent Sessions tracks them in one view no matter which model backs each run. Your keys and data stay yours; the desktop app is local-first.
Can I run several agents at the same time and still keep track?
That is the point of the view. You can launch multiple long-running agents at once, and the overview shows all of them side by side with their working, waiting, or done state, so you never have to open each conversation just to know where things stand.