MULTI-AGENT SYSTEM
Agents, skills and sandboxes,
composed into repeatable work
Create sessions, configure models, attach skills and run sandboxes in one workspace. The left menu is the whole product — set it up once, reuse it everywhere.
- No credit card
- Bring your own key
- 20,000 free tokens a day
Capabilities
Six things, one entry point
- 01
Agents
Bind a role, prompts, tools and skills once; pick the agent in any session.
- 02
Skills
Package a workflow as a skill so the agent loads it on demand.
- 03
MCP
Connect external tools and data sources, managed centrally and composed per agent.
- 04
Sandbox
Every command runs in an isolated environment you can preview and discard.
- 05
Exchange Space
Inject secrets and files into the runtime without putting them in prompts.
- 06
GitHub
Repos, tokens and webhooks in one place, so agents can work on code directly.
Getting started
Three steps to your first session
- 1
Configure a model
Open LLM in the left menu and add your key and model ID — or just use the public pool.
- 2
Start a session
Pick an agent and start chatting; the default config works too.
- 3
Extend as needed
Attach MCP servers, skills and tools, or run commands in a sandbox.
Pricing
Both tiers are $0 — start on the public pool
The public pool gives you 20,000 tokens a day, enough for one real task. When that is not enough, bring your own key: unlimited usage, billed by your provider.
How billing works →Blog
Trade-offs, written down
Your first real task in five minutes
From sign-up to an agent running tests inside a sandbox: four clicks, one paste, plus the three most common ways it fails.
Five trade-offs from building the runtime
Why isolation lives in the database, why streaming uses SSE, why sandboxes are disposable, and where we chose slower but explainable.
Why build another agent workspace
Chat boxes are everywhere. What is missing is a runtime that can finish a task — sessions, tools, sandboxes, skills and MCP without a human ferrying context around.
Docs follow the shipped behaviour; UI changes are recorded in the changelog.