Submit projectSubmit

Sage

ZHangZHengEric / Sage

An open-source agent runtime with Desktop, Server and Python SDK entry points for executing tasks, evaluating changes and retaining candidate versions.

MCPEvaluation & Benchmarks

Save privately. Follow for reviewed updates in your SOTA inbox. Neither changes the ranking.

Your workspace

About Sage

Teams need comparable evaluation records and version history before deciding whether to adopt changes to an agent package.

Who it’s for

  • Python developers building agents with model, tool and Skill integrations
  • Teams comparing feedback-driven changes to agent instructions and packages

When to consider it

Consider Sage when you want task execution and package evaluation in one runtime, with candidate history and explicit version adoption. Its documented RSI workflow separates task execution, evaluation and tuning; general agent integrations include Skills and MCP.

Tradeoffs & limitations

  • The MIT license covers the software; model usage and self-hosted infrastructure can incur separate costs.
  • Standalone Python requires 3.12 or later. Server currently supports one worker; persistence does not establish distributed job takeover.
  • RSI support varies by host. Desktop targets simple agents and native tools; general MCP support does not imply arbitrary MCP tools work in RSI.
  • Local process execution is not container isolation. Hosts must enforce tool permissions, credentials and resource budgets.
  • macOS installers are not notarized and Windows installers are not code-signed.
  • RSI changes agent packages, not model weights. Evaluation scores do not establish general improvement or remove the need to inspect regressions.

A useful first evaluation

Documentation reviewed on November 18, 2026; no installation, agent run, benchmark or security test was performed. Before adopting a candidate, compare it with a baseline on your own held-out tasks, inspect per-case regressions and costs, and confirm the host supports the required tools.

Suggested evaluation, not a report of tests we ran. Download the worksheet.

SOTA overview · Documentation-based assessment · Sources & review method

Updates

No updates shared yet.

Discussion

Newest first

Ask a question or share how you use Sage.

Keep it helpful. Community rules

Loading discussion…

Sources & review method

Documentation-based assessment · Nov 18, 2026 · Prepared with AI assistance; not a hands-on benchmark.

Checked by SOTA · AI-assisted documentation review (interactive Codex). Selection advice is our assessment; verify current requirements for your deployment.

Editorial policy · Suggest a correction

Import history & original evidence

Community submission #25

Scope: Community submission · Imported Nov 18, 2026

An open-source agent harness with Desktop, Server and Python SDK for running agents and comparing feedback-driven candidate versions before user adoption.

Submitted by the repository owner (matching GitHub account ID 45086558). Claims were independently checked against the pinned official documentation; no hands-on test, benchmark or Maker verification is claimed.

ZHangZHengEric/Sage

Scope: Official repository documentation · Imported Nov 18, 2026 · MIT

An open-source agent runtime with Desktop, Server and Python SDK entry points for executing tasks, evaluating changes and retaining candidate versions.

Based on the pinned README, English documentation overview, RSI workflow, support boundaries, release notes and MIT license checked on 2026-11-18. Viewed the official Desktop brand mark and repository workbench screenshot for identity only; no current-version UI behavior, performance or security validation is claimed.