Skip to main content
TELOS is built by LEAP Lab @ Tsinghua University — a research group focused on machine learning, multimodal learning, and embodied intelligence.

Thesis

TELOS makes exactly one claim: context is yours, agents are hired. TelosIR is an engine-agnostic, serializable, portable context representation — your persona, your tools, your 20-turn mid-session thread, all packed onto the same stone tablet. Hand it to Claude today, move it to DeepSeek tomorrow, run it on a local vLLM tonight.

Roadmap

PhaseThesis
Phase 1 · Protocol correctness hardeningTurn “cache cannot be invalidated” from a slogan into a CI red/green light
Phase 2 · Production reliability & observabilityMake the gateway safe to leave on someone else’s prod traffic
Phase 3 · Take over the call chainGo from prompt rewriter to the agent’s traffic plane
Phase 4 · Context becomes an assetTrajectories are no longer logs — they’re forkable code

License

The core is Apache 2.0.

Contributors

Zheng Wang · Shenzhi Wang · HongTao Zhong · Shiji Song · Gao Huang

Citation

Cite the protocol:
@misc{wang2026telos-agent,
  title        = {Telos: A Cost-Aware Inference Infrastructure for AI Agent},
  author       = {Zheng Wang, Shenzhi Wang, HongTao Zhong, Shiji Song, Gao Huang},
  howpublished = {\url{https://github.com/learningCatHD/telos-sdk.git}},
  year         = {2026}
}
Cite the SWE-bench Verified A/B study:
@misc{wang2026telos-swebench,
  title  = {Without Trading Resolution for Tokens:
            A Pre-Registered SWE-bench Verified A/B Study of the TELOS Protocol},
  author = {Wang, Zheng and Wang, Shenzhi and Zhong, HongTao and
            Song, Shiji and Huang, Gao},
  year   = {2026},
  howpublished = {LEAP Lab Technical Report, Tsinghua University},
  url    = {https://github.com/learningCatHD/telos-sdk}
}

GitHub

Star the repo, file issues, read the source.

LEAP Lab

The research group behind TELOS.