Runs, evaluates and improves AI agents from task feedback, keeping each version with its evidence, via a desktop app, a self-hosted server API or a Python SDK.
Use it now
Keep it with up to 9 other tools you're setting up, then share them all as one link.
Try this first
Ask your coding agent:
- Improve agents with RSI
Use Sage to run an agent: ‹what you need›
- Compose tools and MCP
Use Sage to combine native tools, reusable Skills and MCP services: ‹what you need›
- Package agent versions
Use Sage to define an agent's instructions, Skills, tools and resources in sage.yaml: ‹what you need›
Call the API
Auth A bearer token from ZHangZHengEric
Setup guide: github.com/ZHangZHengEric/Sage/blob/main/app/v2/server/READ…
Compare
Press Compare to add it to the tray at the bottom of the window, then pick up to 2 others anywhere on the site.
- Access
- API
- Auth
- Bearer Token
- Skills
- 5
Skills
Improve agents with RSI
Runs an agent on a goal or curated usage history, grades each run with an independent evaluator and keeps tuned versions for you to adopt.
Package agent versions
Defines an agent's instructions, Skills, tools and resources in sage.yaml and keeps each candidate version with its diff and evaluation record.
Compose tools and MCP
Combines native tools, reusable Skills and MCP services in one agent, with the host managing authorization and tool policies.
Pause and resume runs
Persists sessions, streams output, and pauses a run for human input or approval before resuming it.
Serve agents over A2A
Answers other agents over the A2A protocol at /a2a/v1 and serves an Agent Card, with Bearer API keys scoped to read or invoke.
