StartupRemote
OpenRouter
One API for every model — routing, pricing and fallbacks.
OpenRouter is the unified API for frontier and open models: one endpoint, per-token billing across providers, automatic fallbacks and routing by price, latency or capability. Bootstrapped and profitable, it became the default way provider-agnostic harnesses reach every model.
Pros & cons
Each point cites its sources and says what kind of evidence backs it. A second reviewer tried to refute every claim before it was kept.
Pros
- DocsOne API key reaches models from many providers, and if a provider returns an error OpenRouter automatically tries the next one.[1]
- DocsPer-token inference is passed through at provider prices with no markup; the fee is taken when you buy credits.[1][2]
- DocsPrompts and completions are not logged by default (logging is opt-in), and you can exclude providers that log or train on data.[1]
- DocsPay-as-you-go accounts get a $25,000 monthly BYOK allowance with no fee; usage above it costs 5% of the normal OpenRouter price.[1]
Cons
- DocsCredit purchases carry a 5.5% fee ($0.80 minimum), or 5% for crypto, on top of model prices.[1]
- User reportsUsers report that some third-party providers serve quantized models with noticeably worse output, and that you have to pin providers yourself to avoid it.[1][2][3]
- User reportsUsers report credits running out faster than buying direct ($100 lasting two days vs over a week on DeepSeek), which they attribute to cheap providers with short cache TTLs.[1][2]
- DocsOpenRouter reserves the right to expire unused credits one year after purchase, and free models are capped at 50 requests a day until you buy at least 10 credits.[1]
- Best for
- Prototyping and multi-model harnesses that need one key, fast access to new models and automatic provider fallback.
- Watch out for
- Pin providers and quantization levels for production, because the cheapest route may be a degraded or low-cache host.
Assessed Oct 29, 2026 · adversarially reviewedSuggest a correction
People
Co-founder & CEO
Harness engineering
Open roles (28)
- Multimodal, Developer Relations
Remote (US) · Full-time · Posted Nov 19, 2026
EngineeringRemote - Trust & Safety Lead
Remote (US) · Full-time · Posted Nov 19, 2026
OperationsRemote - Software Engineer, Developer Products
Remote (US) · Full-time · Posted Nov 3, 2026
EngineeringRemote - Product Manager, Provider Marketplace
Remote (US) · Full-time · Posted Oct 29, 2026
Product & designRemote - Enterprise Support Engineer (West Coast)
Remote (US) · Full-time · Posted Oct 27, 2026
EngineeringRemote
On X
No posts from them in our X feed yet.
Videos

IndyDevDan · Oct 5, 2026
OpenRouter State Of Models: Jev, Open-Weights, and Tokenomics
Cost & limits · Comparisons & reviews
In research (3)
- Do LLM Agents Still Need to Be Told How to Self-Improve?Verdict: Useful
Nov 17, 2026 · API route used to access the non-Claude models in the Pi harness.
- kw-bench: Breaking Frontier Harnesses on Real-World TasksVerdict: Useful
Oct 26, 2026 · All DeepSeek runs were routed through OpenRouter, including an Anthropic-compatible endpoint for Claude Code.
- EvoRouter: Cost-Efficient Multi-Agent Routing via Insight-Guided On-Policy DistillationVerdict: Incremental
Oct 26, 2026 · Commercial per-request model routing; EvoRouter learns routing over whole multi-agent configurations instead of single models.
Neighbours in the directory
New York, NY · Est. 2021
Serverless cloud built for AI and agentic compute.
Modal lets teams run inference, fine-tuning and arbitrary agent workloads as serverless Python without managing infrastructure — a common substrate for harness backends at scale.
Seattle, WA · Est. 2019
Durable execution for long-running agent workflows.
Temporal's durable execution engine is widely adopted to make agent pipelines reliable: state survives crashes, retries are automatic, and workflows run for minutes or months. A key piece of serious harness infrastructure.
$430M total · $300M Series D (Feb 2026) at $5B valuation
San Francisco, CA · Est. 2021
Durable functions and workflows for AI agents.
Inngest runs agent logic as durable steps — queueing, retries, concurrency control and human-in-the-loop waits — without operating your own workflow infrastructure.
Connections
Relations from the Harness Hub knowledge graph, as charted in the State of Harness reports.
Mentioned in · 1
- Build Day: OpenRouter x MetaEvent · 2026-11-09