OpenAGI
A Python package for structuring, sharing, and retrieving agents built for AIOS.
- Source repo
- agiresearch/OpenAGI
- Stars
- ★ 2.3k
- Last updated
- 1y ago
- License
- MIT
- Primary language
- Python
- FA score
- 18/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specific
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A Python developer building a custom agent for AIOS who needs the repository's prescribed author-and-agent directory layout.
- Not a fit if
- Teams needing agents portable beyond the AIOS platform
- Developers who already migrated to Cerebrum per the README
- Teams needing documented auth, endpoints, or error handling
- Source review
- 18/100 · Major gaps 8 safety controls not found
What does this agent do, and when should you use it?
OpenAGI is the pyopenagi package for creating agents intended for AIOS. It defines an agent layout under pyopenagi/agents, organized by author and agent name, with agent.py, config.json, and meta_requirements.txt. Developers can configure external tools and add new tool files under pyopenagi/tools/. Its interact.py command supports uploading a developed and tested agent or downloading an implementation created by someone else. The README also advises developers building agents for AIOS to migrate to Cerebrum, identified as the latest SDK for connecting with the AIOS kernel.
A developer creates pyopenagi/agents/<author>/<agent_name>/ and places agent.py for execution logic, config.json for configuration, and meta_requirements.txt for dependencies in that folder. External tools can be configured following tools.md, while new tools can be added as files under pyopenagi/tools/. After development and testing, python pyopenagi/agents/interact.py --mode upload --agent <author_name/agent_name> uploads an agent for sharing; the --agent value must exactly match the local folder path. Using the same script with --mode download downloads another developer's named agent implementation.
- A Python developer building a custom agent for AIOS who needs the repository's prescribed author-and-agent directory layout.
- An engineering team separating an agent's execution code, configuration, and dependency declaration into agent.py, config.json, and meta_requirements.txt.
- A developer whose in-progress agent needs external tools and can follow the documented tool setup path.
- A maintainer extending an agent with a new tool file in pyopenagi/tools/.
- An agent author who has finished testing and wants to share a local implementation through interact.py upload mode.
- A developer who wants to retrieve another contributor's implementation using interact.py download mode.
How do you install or deploy this agent?
Install from PyPI:
pip install pyopenagiInstall locally:
git clone https://agiresearch/OpenAGI
cd OpenAGIpip install -e .
The supplied material does not specify a Python version, authentication method, upload service endpoint, or credentials for external tools.
How do you use this agent?
Create pyopenagi/agents/<author>/<agent_name>/ and include agent.py, config.json, and meta_requirements.txt. Upload a developed and tested agent with: python pyopenagi/agents/interact.py --mode upload --agent <author_name/agent_name>. Download an agent with: python pyopenagi/agents/interact.py --mode download --agent <author_name/agent_name>. The README states that the agent argument must exactly match the local folder path.
What are this agent's strengths and limitations?
- Defines a concrete separation of execution logic, configuration, and dependency metadata for each agent.
- Uses one interact.py command interface for both uploading and downloading agents.
- Provides an extension location for new tools in pyopenagi/tools/ and a documented path for external-tool setup.
- Supports both PyPI installation and editable local installation with pip.
- The documented purpose is building agents for AIOS, with no evidence of portability to other platforms or runtimes.
- The README recommends migrating AIOS agent development to Cerebrum, which may require an adoption or migration plan.
- The supplied material does not document Python-version support, upload/download authentication, service endpoints, or error handling.
- It does not enumerate supported external tools, their credentials, or their runtime dependencies.
How does this agent compare with similar options?
For connecting with the AIOS kernel, the README identifies Cerebrum as the latest SDK and advises migration to it for building agents in AIOS.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| OpenAGI This agent | 18 · Major gaps | CLIFree | ★ 2.3k | 1y ago | Python | — |
| LightAgent | 60 · Some gaps | Library / SDKFree + model costs | ★ 1.2k | 13d ago | Python | OpenAI API |
| Langroid | 70 · Some gaps | Library / SDKFree + model costs | ★ 4.1k | 4d ago | Python | Claude Code · OpenAI API |
| AgenticX Multi-Agent Platform | 64 · Some gaps | CLIFree + model costs | ★ 235 | today | Python | Claude Code · OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: The repository does not provide explicit mechanisms for permission management, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, or source attribution. Therefore, all trust criteria score 0. Deduction reason: lack of documentation or implementation in these areas.
Evidence shows: Descriptions in README and pyproject.toml are consistent, but there are inconsistencies such as incorrect clone URL in README (agiresearch/OpenAGI vs actual repository). Dependencies are listed in requirements.txt but without version pinning or integrity checks. Failure messages are not documented. Thus self-consistency scores 1, dependency availability scores 1, failure messages scores 0.
Evidence shows: README describes target audience (AIOS users) and scenarios (creating agents). Capability boundaries are not explicitly defined, but guidelines for adding new agents are provided. Trigger precision is not addressed. Environment fit is partially covered with installation instructions, but no specific OS or Python version requirements are mentioned. Therefore audience and scenarios score 1, capability boundaries score 1, trigger precision scores 0, environment fit scores 1.
Evidence shows: Information architecture is clear with README, LICENSE, pyproject.toml, etc. Installation instructions exist, but the local installation command has a URL error. Naming stability is not explicit, but version numbers exist. Examples and FAQ section has basic examples but no FAQ. Known limitations are not mentioned. License is MIT, but copyright year is 2024, which may not match repository creation time. Versioning changelog is missing. Maintenance responsibility is not explicit, but contribution guidelines exist. Therefore information architecture scores 1, install notes scores 2, naming stability scores 1, examples and FAQ scores 1, known limitations scores 0, license scores 2, versioning changelog scores 0, maintenance responsibility scores 1.
Evidence shows: Output usability is not explicit, but usage examples are provided. Marginal value lies in providing an agent creation framework, but no comparison with alternatives. Cost-benefit is not discussed. Therefore output usability scores 1, marginal value scores 1, cost-benefit scores 1.
Evidence shows: README cites a paper, but no specific implementation details are provided. Cross-source verification is insufficient as only one source is available. Facts and inferences are not clearly separated. Therefore claim traceability scores 1, cross-source corroboration scores 0, fact-inference separation scores 0.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- The clone URL in README is incorrect (agiresearch/OpenAGI vs actual repository), which may cause installation failures for users.
- Dependencies are not pinned, posing supply chain risks.
- No security-related documentation, such as permission management, data flow transparency, etc.
- No known limitations or version changelog, affecting user assessment.