HumanLayer
Getting AI Coding Agents to solve hard problems in complex codebases. HumanLayer lets software contact humans for feedback, input, and approvals, designed for implementing manual approval steps for software- and human-driven workflows, oversight of autonomous AI agents.
At a Glance
- Technology sector
- AI/ML companies building autonomous agents
- Developer tools companies
- Enterprise engineering organizations
- +7 more
AI Tools by HumanLayer
(1)HumanLayer
Human in the Loop Agent SDK
Discussions
No discussions yet
Be the first to start a discussion about HumanLayer
Latest News
The four startups from YC's Fall batch that enterprises should pay attention to
Launch HN: Human Layer (YC F24) – Human-in-the-Loop API for AI Systems
Let's chat about vibe coding & Ralph! (Chat with Dexter at Humanlayer)
Deep dive on Ralph with Geoff Huntley discussing history, cursed lang, and implementation comparisons
Products & Services
An API and SDK that enables tool-calling AI Agents to contact humans for help, feedback, and approvals. Provides Python and TypeScript SDKs for implementing manual approval steps in software- and human-driven workflows, with oversight of autonomous AI agents. Routes requests to the right people through preferred channels (Slack, email, SMS coming soon, Teams coming soon), manages state while waiting for responses, and provides complete audit trail. Works with any AI framework (LangChain, CrewAI, LlamaIndex, etc.) and any language model that supports tool calling (OpenAI, Llama, Claude, etc.). Framework and model agnostic with no lock-in.
An IDE that orchestrates AI coding agents with battle-tested workflows enabling AI to solve hard problems in large, complex codebases. Built on Claude Code. Features include: Superhuman for Claude Code with keyboard-first workflows for builders who value speed and control; Advanced Context Engineering to scale AI-first dev to entire teams; MULTICLAUDE to run Claude Code sessions in parallel with worktrees and remote cloud workers. Full-featured interface for AI-assisted development. Currently in free private beta with many happy users.
An autonomous coding agent implementation using a bash loop structure. Originally presented by Geoff Huntley in June 2025, Ralph enables overnight autonomous coding with subagents. Features a compiler stack originally written in Rust and the programming language 'cursed lang' that Ralph built. Anthropic released an official Ralph Wiggum plugin in December 2025.
A system to require and manage human approvals for AI agent function calls, including multi-team coordination, escalations, and timeouts. Enables safe access for AI while keeping humans in the loop for critical workflow steps.
Market Position
HumanLayer positions itself as framework-agnostic and model-agnostic human-in-the-loop infrastructure, differentiating from competitors like Vista Equity Partners, Mistral AI, and Writer. Unlike traditional agent frameworks (LangChain, CrewAI, LlamaIndex, AutoGen), HumanLayer focuses specifically on the human oversight layer rather than the agent orchestration layer. The platform integrates seamlessly with all major AI frameworks and models, providing a 'happy medium' that brings in human oversight only when needed without slowing down processes. TechCrunch featured HumanLayer as one of four startups from YC's Fall 2024 batch that enterprises should pay attention to, highlighting its role in preventing AI agents from going off the rails while maintaining efficiency.
Leadership
Founders
Dexter Horthy
Started coding at 17 at NASA JPL (Jet Propulsion Laboratory). Spent 7 years as an Engineer, SE, PM, and Exec at Replicated, a series C dev tools startup. Worked on container orchestrators and helped teams like Hashicorp, DataStax, and H2O.ai deliver and ship on-prem k8s products. Previous roles include: VP Solutions Engineering and Customer Success at Replicated (2020-2022), Senior Product Manager/Advisor to CEO at Replicated (2022-2023), CEO and Co-Founder at metalytics.dev (2023-2024), Platform Engineering Lead and Backend Engineer at Aspiration.com (2016-2017), Software Engineer at Sprout Social (2013-2016), Summer Intern at Jet Propulsion Laboratory (2008), and Technical Office Assistant at University of Chicago (2012-2013). Currently Advisor at Prefect since 2021.
Executive Team
Dexter Horthy
Founder and CEO
Started coding at 17 at NASA JPL. Spent 7 years in various roles at Replicated (series C dev tools startup) including VP Solutions Engineering, Senior PM, and Advisor to CEO. Founded metalytics.dev (2023-2024). Experience at Aspiration.com and Sprout Social. Advisor at Prefect since 2021.
Founding Story
HumanLayer started when the founder was building AI agents for data teams. They wanted to automate tedious tasks like dropping unused tables, but customers were rightfully opposed to giving AI agents direct access to production systems. This led to the creation of HumanLayer as an API that enables safe deployment of autonomous AI systems in production by keeping humans in the loop for critical decisions.
Business Model
Revenue Model
API usage-based pricing with credits system, subscription tiers, and enterprise plans. Free tier for development and small-scale projects, pay-as-you-go for production usage, and enterprise plans with whitelabeling, additional features, and priority support.
Pricing Tiers
Includes 1,000 operations per month. Great for development, testing, and small-scale projects.
Credits-based pricing for usage beyond the free tier. Flexible for production usage.
Higher usage caps, additional channels (Microsoft Teams, SMS), private VPC or on-prem deployment options, enterprise features like RBAC/SSO, whitelabeling, additional features, and priority support for teams building customer-facing agents.
Currently free to use during the private beta phase.
Target Markets
- Technology sector
- AI/ML companies building autonomous agents
- Developer tools companies
- Enterprise engineering organizations
- Data teams
- DevOps teams
- Internal automations requiring approval
- Dynamic data labeling
- Customer-facing control and visibility
- AI SDR that drafts personalized sales emails with human approval in Slack
- Ask in Slack before dropping database tables
- AI newsletter where subscribers have email conversations with content
- Boundary
- Syntax Cinema - Tom Granot
- Casco
- Two YC customers with 50-100 person engineering teams