SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
OpenRouter is the leading AI routing and infrastructure layer that enterprises use to access, manage, and optimize large language models across providers without lock-in or unnecessary cost. The company powers advanced AI teams by providing flexibility to scale confidently as models evolve.
You'll join the Ori team, a small group building client-side products for enterprise adoption of OpenRouter beyond engineering organizations. Ori includes a CLI, Slack integration, and desktop application designed to put any model to work in users' terminals and workflows. The team was built the current product in a few months and is led by one of OpenRouter's co-founders. This is a hands-on individual contributor role with broad surface area and minimal process between you and shipping.
Key responsibilities include:
- Building and shipping across all Ori surfaces: the CLI and its harness integrations
- Making model fungibility real by enabling model switching, provider fallbacks, tool and prompt settings tuned to each model, and organizational guardrails without extra authentication
- Building the agent runtime, including sandboxed agents that never see their own credentials, secrets management, session management and compaction, and skills that agents can learn and share across models
- Taking agents from laptop to cloud: deploying, forking, and shutting down agents on command with version-controlled and auditable agent state
- Owning quality end-to-end from install and auth through CLI UX, reliability, observability, and evals. Ori is heavily dogfooded internally
- Feeding gaps Ori exposes back into OpenRouter's API and SDKs so the ecosystem can build multi-model agents on shared primitives
You'll work alongside product, design, and platform teams, with direct collaboration on a small team with fluid boundaries.
REQUIREMENTS:
- 5+ years shipping production software with real ownership of developer tools, CLIs, agent products, or infrastructure behind them
- Daily fluency with coding agents (Claude Code, Codex, OpenCode, Pi, Hermes) and specific opinions about what makes agent loops, harnesses, and tool integrations effective
- Solid grasp of how agents work under the hood: tool calling, context management, prompt caching, evals, and cost/latency/quality tradeoffs between models
- Good systems judgment for running agents safely: sandboxing, process isolation, secrets management, and reliability across many users
- High agency and bias toward action; comfortable talking to users, prototyping, shipping, and iterating on a small team
- Product taste for developer tools; attention to interaction details and how the first five minutes feel
NICE TO HAVE:
- Built or maintained an open-source agent harness, CLI, or developer tool
- Experience with multi-agent systems, agent skills or memory, or eval and benchmarking tools
- Shipped a Slack app or native macOS app
- Existing user of OpenRouter or Ori, or active side projects in AI tooling
