HarnessRouter
HarnessRouter is a single API that brings leading AI agents like Codex, Claude Code, and Hermes into your app, handling the full agent backend so you can ship AI features in minutes.
At a Glance
About HarnessRouter
HarnessRouter is a Y Combinator-backed API platform that lets developers embed production-grade AI agents—Codex, Claude Code, Hermes, and more—directly into their products without building or maintaining the underlying agent infrastructure. Instead of spending months wiring up sandboxes, runtimes, tool orchestration, and streaming, developers get a single API key and a drop-in AGENTS.md guide that a coding agent can use to wire the integration automatically.
What It Is
HarnessRouter sits between your product and the world's leading AI agents. It provides what the company calls a "harness"—the full execution layer around a model, including instructions, model policy, tools and MCP, skills, runtime and sandbox, permissions and guardrails, and an output contract. The result is that your users receive finished artifacts (videos, codebases, documents, images, game files) rather than raw model text, all without leaving your product.
How the Three-Step Workflow Works
The integration path is designed to be completed by a coding agent, not manually:
- Step 1 – Drop in AGENTS.md: Copy the official HarnessRouter instructions and paste them into Codex, Claude Code, or another coding agent.
- Step 2 – Describe your feature: Tell the coding agent what you want to build (e.g., "Build a contract-review app that lets users upload an agreement, streams progress, and renders the result").
- Step 3 – Add your API key securely: When the coding agent opens a secure secret modal, create a Workspace API key and paste it in. The agent stores it as
HR_API_KEYand completes the build.
The demos on the site were built exactly this way, according to the product page.
Harness Architecture and Agent Roster
Each harness is a named, swappable configuration. Your integration code and output schema stay constant; switching from claude-code to codex requires only a one-line config change. Current harnesses include:
- Codex – Code, apps, and images
- Claude Code – Code, files, and documents
- Hermes – Autonomous tasks across any frontier model
- Pi – Coding agent (listed as "Soon")
The platform handles sandboxing per run, agent runtime, tool orchestration, files and artifacts, sessions and streaming, retries and timeouts, permissions, and cost controls—the full list of infrastructure concerns the product page describes as taking months to build in-house.
Control and Observability
HarnessRouter provides a trace view that shows every task broken down step by step, including event timelines with timestamps, tool calls, file changes, and final result artifacts. Operators can configure each harness individually and set budgets, alerts, and hard caps so production usage stops at a defined limit.
Why It Got Attention
HarnessRouter was ranked #5 Product of the Day on Product Hunt at launch. The product is backed by Y Combinator and lists a launch partner (TokenRouter). The homepage lists several organizations as users, including Epsilla, Hibo, Readily, Spira AI, and Stanford Medicine—these are vendor-published claims on the product page.
Update: Active Launch Phase
The product is in active launch mode with a 7-day card-gated trial available. The pricing page notes that checkout shows $0 today with the renewal date and selected plan visible before commitment. The platform roadmap includes at least one additional harness (Pi) listed as "Soon," signaling ongoing agent roster expansion.
Community Discussions
Be the first to start a conversation about HarnessRouter
Share your experience with HarnessRouter, ask questions, or help others learn from your insights.
Pricing
7-Day Free Trial
Card-gated trial. Checkout shows $0 today with renewal date and selected plan visible before commitment.
- Access to platform
- Production credits included
- Budgets and hard caps
Developer
For building and launching your first AI app.
- Platform subscription with credits for production runs
- Top up production usage when you need more
- Set budgets and hard caps
Production
For running a production AI app.
- Platform subscription with credits for production runs
- Top up production usage when you need more
- Set budgets and hard caps
Scale
For scaling production AI apps.
- Platform subscription with credits for production runs
- Top up production usage when you need more
- Set budgets and hard caps
Enterprise
Annual contracts, committed usage, private deployment, security review, audit controls, and SLA.
- Annual contracts
- Committed usage
- Private deployment
- Security review
- Audit controls
- SLA
Capabilities
Key Features
- Single API for multiple AI agents (Codex, Claude Code, Hermes)
- Drop-in AGENTS.md for coding agent integration
- Managed agent runtime and sandbox per run
- Tool orchestration and MCP support
- Files and artifacts output (videos, code, documents, images)
- Sessions and streaming
- Retries and timeouts
- Permissions and guardrails
- Cost controls with budgets and hard caps
- Step-by-step task tracing and event timeline
- Swappable harness configuration (one-line switch)
- Structured renderable results (diffs, files, images)
- Workspace API key management
- Usage Wallet with top-up model
