Honest comparison, updated October 2026
HumanLayer today is an AI coding IDE for orchestrating agent sessions at team scale, and its own README calls its classic approvals SDK deprecated. Pushary asks a person while your agent runs. That person can be you or one of your own users, and they answer on their phone. When the question goes to your users, no answer means no, and every answer is saved.
These two used to compete on approvals. Mostly they no longer do. HumanLayer's job now is supervising how agent fleets build software: parallel sessions, worktrees, and human review of research and design documents inside its IDE. Pushary's job is the decision while an agent runs. An agent, yours or the one you ship to customers, stops at a risky step. A specific person gets the question on their phone, and the agent goes ahead only on a yes.
require_approval, human_as_tool, and the Slack and email contact channels belong to the old HumanLayer monorepo. Its README says the code there is pretty much all deprecated, and the old approval doc pages return 404. If that workflow brought you here, Pushary is the maintained way to run it now. decisions.ask() waits for a real answer, treats no answer as no, and keeps the record. It is live and self-serve today.
Pushary is built so the agent you ship can ask its own users, each about their own action, at scale. Each decisions.ask() reaches exactly that person, under your product's name. HumanLayer's current product is a workspace for your own developers.
HumanLayer's oversight happens in its own app, on web, desktop or phone, for the sessions it runs. Pushary sends the question to the lock screen of whoever must decide, for agents you already run. Yes/no questions are answered right there, and a choice or a typed answer opens the app. Rules per tool decide what needs a person at all.
For orchestrating fleets of coding agents, HumanLayer is a serious product from a team with a strong track record (they authored 12-factor agents and influential context-engineering work). Parallel sessions, worktrees, daemons, and review of plans before code exists catch a bad direction earlier than an approval on a single tool call can. Bring-your-own-key pricing with a free starter tier makes it cheap to evaluate. If you want one app to run your team's coding agents, evaluate HumanLayer for that job. Pushary does not orchestrate sessions, and does not try.
| Feature | Pushary | HumanLayer |
|---|---|---|
| Maintained ask-a-person API you can use todayCORE | ||
| Your agent asks your own usersUSERS | ||
| Your users connect their phone with a link | ||
| Approve or Deny on your phone's lock screen | ||
| No answer means no, for questions to your users | ||
| Every decision saved, searchable and exportable | ||
| Rules per tool, changed with one API call, no deploy | ||
| Adds approvals to agents you already run: hooks for Claude Code, Codex, Cursor, Gemini CLI and Hermes; MCP tools for any MCP client | ||
| Human review checkpoints on agent design artifacts | ||
| IDE orchestration of parallel coding-agent sessions | ||
| Multi-repo worktrees and cloud daemons for agent fleets | ||
| Pricing | $9.99/mo Agent, $99/mo Partner, self-serve | Free for up to 3 members, Pro $100 per user a month |
Competitor rows reflect HumanLayer's public site and repository state as of October 2026. Their product is evolving quickly, so check their site for current state.
As of October 2026, HumanLayer is an AI coding IDE and collaboration platform for orchestrating coding-agent sessions at team scale. It runs parallel Claude Code and Codex sessions with your own keys, multi-repo worktrees, and local and cloud daemons. Its structured workflow puts human review on research and design documents before implementation. It went through a CodeLayer rebrand in late 2025, has since returned to the HumanLayer name, and has public self-serve pricing.
The classic SDK (require_approval, human_as_tool, Slack and email contact channels) is deprecated. The company's own GitHub README says the code in the old monorepo is, in their words, pretty much all deprecated. The old approval docs pages return 404, and the Python package has not shipped since mid-2025. There was no shutdown announcement, so it is fair to call it unmaintained rather than shut down. It is not what their current product markets.
They now solve different problems. HumanLayer supervises how agent fleets build software inside its IDE, with people reviewing plans and designs. Pushary asks a person while an agent runs. When an agent reaches a step a person must decide, decisions.ask() sends the question to that person's phone and waits for the answer. No answer means no, and the outcome is saved. If you came here looking for the old require_approval workflow, Pushary is the maintained way to do that job today.
With Pushary, yes. That is what it is built for. You connect each user's phone once with a link, tied to your own user id. Each decisions.ask() then reaches that specific user about their own action, and the notification and the decision page show your product's name. HumanLayer's current product is for your own development team. HumanLayer publishes no public docs, and its site describes no way to connect your users or a decisions API.
Pushary is self-serve. The Agent plan is $9.99/mo for a solo developer approving their own agents, and Partner plans start at $99/mo for products whose agents ask their own users. HumanLayer is free for up to 3 members, and Pro is $100 per user a month, as of October 2026. Check their pricing page for current numbers.
HumanLayer's IDE orchestrates coding agents such as Claude Code and Codex under your own keys. Pushary adds approvals and notifications to the agents you already run: hooks for Claude Code, Codex, Cursor, Gemini CLI and Hermes; MCP tools for any MCP client. Its SDKs add phone approvals to agents you build with the Vercel AI SDK, LangGraph, LangChain, CrewAI, Mastra, the OpenAI Agents SDK, the Claude Agent SDK, Pydantic AI, Haystack, Eve, Hermes or OpenClaw.
Connect a phone once with a link. Your agent asks and waits for the answer. When the question goes to your users, no answer means no. Every answer is saved. For your own agents and for the ones you ship to your users.
Start free