A single interface to use and evaluate different agent frameworks
1.2k
Stars
94
Forks
26
Open issues
18
Contributors
AI Analysis
any-agent provides a unified interface to run and evaluate agents across multiple frameworks (Google ADK, LangChain, LlamaIndex, OpenAI Agents, Smolagents, Agno, and TinyAgent). It is now in soft deprecation, with core functionality graduating to mozilla-ai-tinyagent; it remains useful specifically for teams that need to compare or run agents across multiple frameworks under one API. Best suited for researchers, framework evaluators, and organizations standardizing on multi-framework agent de...
Inferred from signals mentioned in the README (tests, CI, type safety) — not a review of the actual code.
AI's overall editorial judgment — not an average of the bars above, can weigh other factors too.
Mozilla's abstraction layer for agent framework interoperability, now in soft deprecation
any-agent provides a unified Python API to instantiate, run, and evaluate agents across seven different frameworks (TinyAgent, Google ADK, LangChain, LlamaIndex, OpenAI Agents, Smolagents, Agno). Built by Mozilla AI as a research project to identify common patterns, it has graduated its core insights to mozilla-ai-tinyagent. The project remains actively maintained for multi-framework evaluation scenarios but receives no new feature development.
Launched March 2025 by Mozilla AI as a comparative research effort to distill a minimal, framework-agnostic agent interface. Within ~16 months, the research yielded a leaner core (tinyagent), which is now the recommended path for new projects. any-agent persists as a specialized tool for teams needing to benchmark or switch between frameworks.
Gained 1,183 stars over 16 months (modest slope), with 4 stars in the final week—consistent with a mature, niche tool rather than a viral adoption curve. Growth appears driven by practitioners evaluating multiple agent frameworks and researchers comparing framework capabilities, not by momentum toward becoming a dominant platform.
Adoption not verified. No case studies, testimonials, or disclosed production deployments mentioned in README. Stars (1,183) and forks (94) are modest compared to peer projects (OpenAI Agents: 27,640; microsoft/agent-framework: 11,860). However, adoption not verified does not mean adoption does not exist; niche enterprise or research usage could be undocumented.
Based on README, any-agent wraps seven agent frameworks under a common AgentConfig and AnyAgent interface. Appears to translate user intent into framework-specific calls and normalize output traces. Likely uses adapter/bridge pattern to handle framework heterogeneity. Supports tool abstraction, callbacks, and multi-agent composition (agents-as-tools).
README mentions integration test workflow badges but does not detail coverage percentage or scope. Presence of documentation build and integration test CI suggests active testing discipline, but depth unknown.
Last push 2026-07-01 (3 days before evaluation date), indicating active maintenance. CI/CD badges present. README explicitly states soft deprecation and directs new users to tinyagent—a transparent, honest deprecation posture rather than silent abandonment. This suggests the maintainers are responsive and setting expectations clearly.
ADOPT IF: you need to run identical agent code across multiple frameworks for benchmarking, evaluation, or framework migration testing, and you accept that no new features will be added. AVOID IF: you are building a new agent-based product or need ongoing feature development—migrate to mozilla-ai-tinyagent or choose a framework-specific abstraction instead. MONITOR IF: you are evaluating agent frameworks as a research effort; any-agent's trace/evaluation tooling and multi-framework support remain valuable for that use case despite soft deprecation.
Independent dimensions
Mainstream potential
2/10
Technical importance
5/10
Adoption evidence
2/10
- Soft deprecation means bug fixes only; framework updates (especially newer versions of LangChain, LlamaIndex, OpenAI SDK) may not be tracked, risking integration drift.
- Documentation and cookbooks may become stale relative to framework evolution; users will need to maintain their own adapter code.
- Adoption not verified; community size is small, so vendor support and third-party tooling are unlikely to emerge.
- Multi-framework wrapper adds a layer of abstraction that could obscure framework-specific capabilities and debugging; useful for evaluation, risky for production systems with performance constraints.
- Recommended successor (tinyagent) is separate; migration is required for teams wanting ongoing support, potentially creating churn if tinyagent's API diverges significantly.
any-agent will remain a stable, low-maintenance research artifact. Adoption will stay niche—primarily researchers and framework evaluators. No mainstream adoption is expected; instead, the project serves as a reference implementation of a minimal agent interface (a role it has now ceded to tinyagent). Likely to enter archival or read-only status within 1–2 years if no new framework integrations are requested.
Newsletter
Get analyses like this every Monday
Free weekly digest of the most interesting open-source discoveries.
Languages
Information
- Language
- Python
- License
- Apache-2.0
- Last updated
- 1w ago
- Created
- 16mo ago
- Analyzed with
- anthropic/claude-haiku-4-5
Stars over time
Contributors over time
Top 100 contributors only — repos with more will plateau at 100.
Open issues
Top contributors
Recent releases
Similar repos
mozilla-ai/any-llm
any-llm provides a unified Python interface for communicating with multiple LLM...
microsoft/agent-framework
Microsoft Agent Framework is a production-grade, multi-language framework for...
| Repository | Stars | Week Δ | Language | Score | Updated |
|---|---|---|---|---|---|
|
mozilla-ai/any-agent
this
|
1.2k | +6 | Python | 7/10 | 1w ago |
|
mozilla-ai/any-llm
|
2.1k | — | Python | 8/10 | 1d ago |
 openai/openai-agents-python
|
27.8k | — | Python | 8/10 | 2h ago |
|
microsoft/agent-framework
|
12k | — | Python | 8/10 | 13h ago |
 modelscope/ms-agent
|
4.3k | — | Python | 7/10 | 2d ago |
 jjyaoao/HelloAgents
|
2.3k | — | Python | 7/10 | 1mo ago |
Leaner, recommended successor. Focuses on core agent loop only; any-agent adds multi-framework evaluation. For new projects, tinyagent is the canonical choice; any-agent is only preferred if multi-framework comparison is the explicit goal.
Broader ecosystem play with 11,860 stars. any-agent is framework-agnostic; microsoft/agent-framework is opinionated. Direct replacement unlikely; they serve different design philosophies.
Deep integration with OpenAI's ecosystem (27,640 stars). any-agent includes OpenAI as one of seven options. Users locked into OpenAI's workflow have no need for any-agent; users comparing frameworks do.
Established frameworks with their own abstractions. any-agent wraps them; it does not compete but rather enables cross-framework evaluation of code written for these systems.
Similar multi-framework comparison project (2,258 stars). Differentiation unclear from metadata alone; both appear to target framework evaluation, but any-agent has Mozilla backing and clearer deprecation/migration path.
