← All organizations

AI coding tools and agent collaboration

HumanLayer

HumanLayer builds an AI coding IDE and collaboration platform for engineering teams working in complex codebases. Teams can organize agent sessions, worktrees, research, and implementation plans around shared tasks, with design comments feeding directly back to agents. Engineers bring existing Claude, Codex, or other supported subscriptions and API keys, and run agents locally or in the cloud for background work.

Founded in 2023 by CEO Dexter Horthy, HumanLayer later added Kyle Mistele as cofounder and CTO. Its original human-approval tooling for agents expanded into coding tools, including the CodeLayer private beta in 2025. Its Research/Plan/Implement approach structures coding around codebase research, reviewed plans, and execution, compacting context into reusable artifacts with human checkpoints. Its 12 Factor Agents guide also advocates explicit control over prompts, context, and execution, with small, focused agents and human intervention.

As of August 2026, the company reported that thousands of engineers had used its open-source workflows and that its users ranged from YC startups to Fortune 500 companies. It also reported more than $3 million in cumulative funding from Y Combinator, Pioneer Fund, and angels. Its commercial offering combines a free tier for teams of up to three with paid individual and team plans, plus enterprise options for SSO, audit logs, and private deployment.

www.humanlayer.com

5 talks

Newest first

2 speakers at AIE

Affiliations reflect their AIE appearances, not necessarily current employment.

Start here

  1. 12-Factor Agents: Patterns of reliable LLM applications

    Start with the essay based on Dex Horthy's talk to understand how a mostly deterministic application can introduce model-directed decisions incrementally.

    Dex HorthyAI Engineer World's Fair 2025

  2. No Vibes Allowed: Solving Hard Problems in Complex Codebases

    Read the essay based on Dex Horthy's talk for the role of subagents in compressing investigations and progressive disclosure in loading relevant repository guidance.

    Dex HorthyAI Engineer Code 2025

  3. Loop Engineering from first principles

    Watch Kyle Mistele's talk for an RPC API migration to Effect that tracks progress in version control and divides work into low-risk pull requests.

    Kyle MisteleAI Engineer World's Fair 2026

  4. Harness Engineering is not Enough: Why Software Factories Fail

    HumanLayer's Dex Horthy argues that fully autonomous software factories cannot preserve complex codebases through harness engineering alone: coding agents accelerate implementation while inadequate review, incidents, and architectural decay accumulate.

    Dex HorthyAI Engineer World's Fair 2026

Messages from the stage

Keep model decisions inside explicit software control

Horthy separates a model's structured output from the application code that executes it. His 12-Factor Agents framework treats prompts, orchestration, recoverable state, and human participation as mechanisms developers need to inspect and improve.

Make repository understanding reviewable

Horthy presents intentional context compaction and Research, Plan, Implement workflows as ways to carry useful findings between agent sessions. Targeted research and reviewable Markdown artifacts help keep both agents and humans oriented in established codebases.

Bound autonomous work and verify its effects

Mistele describes coding loops that measure a desired codebase state and make small, reviewable changes; Horthy questions whether harness engineering alone can protect long-term maintainability. The joint loops debate examines competing views of practical autonomy, verification, and software factories.

Affiliations reflect each recorded session, not necessarily current employment.

Company sources · checked 2026-08-27