hey folks! been working HumanLayer for a few months now and stoked to share it with the PH community!
HumanLayer was born from our frustration in trying to safely deploy AI agents in production. We were working on a set of SQL-warehouse-manager agents that would review audit logs and drop unused tables or evolve schemas. We wanted to build agents than went beyond simple recommenders or copilots, we wanted to build autonomous agents that did real work with tool calling.
HumanLayer is an API and SDK that lets AI agents contact humans for feedback, input, and approvals. It's perfect for:
- Implementing manual approval steps for software- and human-driven workflows
- Oversight of autonomous AI agents
- Managing the transition between the two
There are many AI frameworks out there, but none have robust support for human oversight, at least not for agents that run outside of chat interfaces (some call these "outer loop" agents). For these agents, teams have three choices:
1. Hamper these agents, don't give them any tools that are high-stakes, basic deploy agents that are so harmless they don't need human oversight
2. Build complex custom approval workflows that are hard to get right and don't scale
3. Yolo things out and hope for the best
HumanLayer lets you ship high-stakes agents with human oversight, and use that feedback to iteratively improve your agents over time.
- It seamlessly integrates human oversight into any AI workflow, ensuring your agents operate safely while maintaining autonomy.
- It lets your agent's users where they are, whether that's Slack, Email, SMS, or other channels
- It is framework and model agnostic — no lock-in here! Works with any AI framework (LangChain, CrewAI, etc) and any language model (OpenAI, Llama, Claude, etc).
Because HumanLayer provides simple Python and TypeScript SDKs, you can integrate it anywhere. Get started today with a simple:
> pip install humanlayer
And drop a `require_approval()` decorator on any function that needs human oversight.
Want to see it in action? Check out our docs or examples repository for LangChain, CrewAI and other integrations, or join our Discord community
(And we're streaming live answering comments from 12:01am to 1:00am PT at https://www.twitch.tv/dexhorthy - come say hi!)
Congrats! I love how this app flips the typical “AI Agent” script — instead of humans asking for help proactively, agents will be doing stuff all the time and occasionally ping us humans for help when they get into something tricky!
@ihuzaifashoukat thank you!!! I think any industry that needs auditing and compliance can benefit from HumanLayer as a governance layer, although personally with my SRE/DevOps background, I'm *most* excited about use cases for agents that manage infrastructure - now you can go beyond just making recommendations about incident, but the agent can go so far as to perform a scaling operation or restart a pod
HumanLayer
Codebuff
HumanLayer
HumanLayer