dari.dev GTM brain

dari.dev provides model routing for developers running AI agents who want to manage model costs and performance.

category

AI agent model routing

market profile

  • text: Agent teams face a tradeoff between model quality and total task cost, including prompt-cache effects when models switch.
  • text: Dari has shifted from its earlier browser-agent positioning toward routing, so older launch coverage should not be treated as its current offering.

personas

  • title: AI application engineer
    pains:
    • Hard-coding one model across every agent step can waste spend on easier tasks.
    • Routing must preserve task quality and integrate with existing clients.
  • title: Developer productivity or platform lead
    pains:
    • Agent adoption raises token costs across coding workflows.
    • Needs workload-specific evaluations and visibility before changing model selection for the team.
  • title: Coding agent power user
    pains:
    • Wants to continue using Claude Code, Codex or another installed agent while managing model choices.
    • Managing provider credentials and routing setup adds friction.

icp

  • text: Software teams and AI startups operating multi-turn agent workloads where inference spend is material; headcount and geography are unconfirmed.
  • text: Engineering organizations already using Claude Code, Codex, OpenCode or Pi, or building against compatible model APIs.

message pillars

  • text: Use a single endpoint to select among models.
  • text: Keep installed coding agents and launch them through the router.
  • text: Inspect the model serving each turn.

objections

  • objection: Will model selection reduce quality or actually save money?
    response: Benchmark representative tasks against the current model and include cache effects in cost per completed task.
  • objection: Must we replace our current coding agent?
    response: The CLI documents launchers for existing Claude Code, Codex, OpenCode and Pi installations; check compatibility for the deployed version.
  • objection: How predictable is the cost?
    response: Confirm current plan terms directly; independent directory prices conflict and the current pricing page could not be read.

competitors

  • name: OpenRouter
    why: Multi-model gateway and routing alternative for developers consolidating inference access.
  • name: Not Diamond
    why: Model selection service with a coding-agent routing offering.
  • name: LiteLLM
    why: Gateway and routing library teams can use to build their own model selection policies.

influencers

  • name: Avyay Varadarajan
    url: https://www.ycombinator.com/companies/daridev
    why: Founder and CEO.
  • name: Benjamin Hong
    url: https://www.ycombinator.com/companies/daridev
    why: Founder and CTO.

keywords

  • AI model router
  • coding agent token costs
  • Claude Code model routing
  • LLM routing
  • Codex cost optimization
  • agent cost per task

voice

  • Developer-oriented
  • Technical
  • Concise

signals

  • text: An engineering team is scaling coding agents and has begun tracking inference spend per task.
  • text: A team uses several models or agent harnesses and is evaluating routing rather than a fixed model.

stack

  • Claude Code
  • Codex
  • OpenCode
  • Pi
  • OpenAI-compatible API

pricing

model: unknown

motion

product_led

channels

  • GitHub
  • Developer documentation
  • Search

confidence

medium

Founders

  • name: Avyay Varadarajan
    title: Founder/CEO
    role: CEO
    bio: Building dari.dev. Previously CS @ Caltech, SWE @ Uber, ML @ Lakers, startups.
  • name: Benjamin Hong
    title: Founder/CTO
    role: CTO
    bio: Building dari.dev

Back to YC companies