Sage
An extensible framework for custom, self-evolving AI agents, with configurable agent packages, tools, skills, and multi-round recursive self-improvement (RSI).
No outside reports on file
Sage is a powerful but experimental framework that puts safety responsibility on you—use it only in controlled environments with limited permissions until you fully understand how your specific configuration behaves.
Based on public adoption and transparency signals as of Oct 19, 2026. Our opinion, not a code audit or security guarantee.
What it can access
- Local filesNo local folder access notedLOW
- Command executionNo shell command runner notedLOW
- Network reachConnects to outside service APINOTED
- CredentialsNo API token requiredLOW
- Package sourceGitHub (ZHangZHengEric/Sage)PUBLIC
Inferred from this listing’s public package details and description. Check the project’s source before installing.
Install
No npm or PyPI package is listed. Install from source: github.com/ZHangZHengEric/Sage ↗
About Sage
Sage is a framework for building AI agents that can improve themselves over multiple rounds of thinking and action. It lets you customize agent behavior through packages, tools, and skills, with a focus on recursive self-improvement where the agent learns and adapts across conversations.
Who it’s for
Developers and researchers interested in experimenting with self-evolving AI systems and building agents that can refine their own strategies. It's most useful for those comfortable working with custom frameworks rather than ready-to-use services.
Alternatives
- AutoGPT: Open-source agent framework with internet access and memory systems.
- LangChain agents: Flexible framework for building and chaining AI tool use.
- CrewAI: Multi-agent framework focused on task delegation and collaboration.
- OpenAI Assistants API: Managed service for custom AI agents without self-modification complexity.
Setup
Check these steps against the project’s own instructions before running anything.
Follow the project's README. The GitHub repository appears to be the main source; check there for installation instructions specific to your environment.
Why this grade
Signal-based grade (3/7 points): well starred on GitHub, public source code. This reflects adoption and transparency, not a code review.
Popularity and upkeep
- GitHub stars
- 1.2k
- Listing last refreshed
- 2026-10-29
These are public popularity signals, not a maintenance check. Look at the repository’s commit history and open issues.
Security reports
No reports found for this package in our research (last checked 2026-10-29).
Risks to watch for
- Unlimited self-modification: The agent's ability to improve itself recursively could lead to unexpected behavior changes that are hard to predict or undo.
- Tool access scope: Depending on what tools and skills you attach, the agent could potentially access files, APIs, or data beyond your intended scope.
- Input injection: Custom agent packages might not safely validate user inputs, letting attackers craft prompts that override intended behaviors.
- No built-in safeguards: This is a framework you configure; safety depends entirely on how carefully you set it up and what permissions you grant.
- Debugging difficulty: Multi-round self-improvement makes it harder to trace why the agent behaves a certain way or where it went wrong.
How to lower the risk
Least-privilege permissions
- Only attach tools and skills the agent truly needs for its task.
- Restrict file system and API access to specific directories or endpoints.
Network isolation
- Run the agent in a sandboxed environment or container with limited outbound network access.
- Test on non-production data until you're confident in the agent's behavior.
Monitoring
- Log all decisions and self-improvements the agent makes so you can review them later.
- Set up alerts if the agent requests permissions or tools outside expected scope.
Red flags to watch for
- The agent attempting to modify its own rules or bypass restrictions you set.
- Unusual or circular reasoning loops that suggest the self-improvement is not converging.
- Requests for access to sensitive APIs, credentials, or production systems.
Questions
How do I install Sage?
Sage does not publish a standalone npm or PyPI package. Clone or follow the setup instructions in its source repository at github.com/ZHangZHengEric/Sage.
Are there security reports about Sage?
No reports found for this package in our research (last checked 2026-10-29). Sage currently holds Grade C (Moderate risk signals) based on public adoption and transparency signals. Key risk areas to watch include unlimited self-modification, tool access scope, input injection.
Is Sage actively maintained?
Sage has 1.2k GitHub stars. Our directory data was last refreshed on 2026-10-29. Check recent commit activity, release tags, and open issues at github.com/ZHangZHengEric/Sage before pinning it in production.
What are alternatives to Sage?
Notable alternatives to Sage include AutoGPT (Open-source agent framework with internet access and memory systems), LangChain agents (Flexible framework for building and chaining AI tool use), CrewAI (Multi-agent framework focused on task delegation and collaboration), OpenAI Assistants API (Managed service for custom AI agents without self-modification complexity). You can also compare Sage side by side with other MCP servers in the Related tools section on this page.
README badge and team notes
Maintain or use Sage? Show its live grade in your GitHub README, docs site, or run a check from the terminal before you install.
GitHub README Markdown:
[](https://agentarmorhq.com/skills/zhangzhengeric-sage)HTML embed:
<a href="https://agentarmorhq.com/skills/zhangzhengeric-sage"><img src="https://agentarmorhq.com/api/badge/zhangzhengeric-sage" alt="AgentArmorHQ Risk Grade for Sage" height="20" /></a>Terminal check:
curl -fsSL https://agentarmorhq.com/api/cli | node --input-type=module - check "zhangzhengeric-sage"Share with your security lead
Need sign-off to use Sage? Copy this note for Slack, a ticket or a pull request.
### AgentArmorHQ Security Review Memo: Sage
- **Listing URL:** https://agentarmorhq.com/skills/zhangzhengeric-sage
- **Type:** App connector
- **Risk Grade:** Grade C (Moderate risk signals)
- **Outside Security Advisories on File:** 0
- **Public Adoption:** 1.2k GitHub stars
#### Permission & Blast-Radius Signals
- **Local files:** No local folder access noted
- **Command execution:** No shell command runner noted
- **Network reach:** Connects to outside service API
- **Credentials:** No API token required
- **Package source:** GitHub (ZHangZHengEric/Sage)
#### Recommended Hardened Setup
- **Source Repo:** https://github.com/ZHangZHengEric/Sage
- **Controls:** Pin an exact version in `.mcp.json`, scope access to a single workspace folder, and use a least-privilege read-only token where supported.
_Based on public adoption and transparency signals as of Oct 19, 2026. Our opinion, not a code audit or security guarantee._Key facts
- Type
- App connector
- Publisher
- ZHangZHengEric
- Source code
- github.com/ZHangZHengEric/Sage
- Est. context load
- ~900 tokens / turn (~5 tool schemas, 0.45% of 200k context)
- Popularity
- 1.2k GitHub stars
- Last updated
- 2026-10-29
- Last graded
- 2026-10-19
Community
Discussions for Sage
Ask questions or share configuration tips. Governed by our Community Rules & Terms.
No posts match this filter yet.
Be the first to ask a question or share a working tip.
Check before you install, in Claude
Your assistant answers Block, Review or No known issues for any package before it installs it.
Get this check inside Claude