Tool overview
dari.dev is listed under AI Infrastructure & MLOps AI tools.
What is dari.dev?
dari.dev is a drop-in routing layer for AI agents. It exposes OpenAI-compatible endpoints, applies custom rules or a fine-tuned selector model, and routes each request to the least expensive enabled model that meets the configured quality bar.
Best for
AI engineering teams optimizing inference cost across multiple models
Who is it for?
Decision note
Rebuilt from the original export under V412/V411 independent factual verification. Preview only; Apply requires Failures = 0, Warnings = 0, Unmapped = 0, Missing = 0, reviewed explicit clears, image import disabled or V380 PASS, representative edit-screen comparison, and a zero-change post-Apply Preview.
Key features
OpenAI-compatible completions and Responses endpoints
Custom rule-based routing policies
Fine-tuned SLM router for coding-agent traffic
Compatibility with Claude Code, Codex, and Pi
Use cases
Reduce model spend for coding agents
Apply quality-aware provider selection
Centralize model policy across agent clients
Audit real traffic before defining routing rules
Pros
- Drop-in base-URL change for supported clients
- Preserves streaming and tool use
- Supports custom enterprise router training
Cons
- No public numeric price
- Routing quality requires real traffic evaluation
- Model switching can complicate cache economics
Limitations
The current product is a model-routing service, not the older stateful agent-deployment layer described in the stored profile.
Commercial terms and custom router training require direct engagement, and no human-language support claim was published.
Pricing details
Billing options
Custom routing and enterprise agreement
Pricing note
dari.dev uses a sales-led model for routing audits, production traffic, and custom enterprise routers. The official site publishes no stable numeric price or general self-service trial.
Supported languages
- English
Integrations
OpenAI SDKs
OpenAI Responses API
Anthropic gateway
Claude Code
Codex
Pi



Please log in to join the discussion.