angel291592/Intent-Router

Intent compiler for AI agents — converges vague requests into typed IntentSpec contracts (probe, ask, or halt before routing), the input layer for routers and typed-decision models like Jev & Laya

What it solves

It eliminates the "guesswork" and "endless interviewing" common in AI agents. Instead of an agent guessing what you mean or asking you dozens of questions, this tool treats intent like a compiler treats code: it probes existing data (code, docs, tickets) to resolve ambiguities, asks only the one question that truly requires human judgment, and refuses to act if the request is still too vague.

How it works

It processes requests in three passes:

  1. Parse: Converts your request into a draft IntentSpec (a structured plan) and identifies missing information.
  2. Resolve: Classifies every missing piece of information. If the answer exists in your repo, docs, or logs, it probes them automatically. If the answer requires a human decision (like a trade-off or irreversible action), it asks you exactly once.
  3. Typecheck & Emit: It stops if the request is underspecified. If it proceeds, it saves a machine-readable IntentSpec file that acts as a contract for future sessions, ensuring the agent doesn't "forget" your decisions.

Who it’s for

Developers and teams who want to build more reliable AI agents that respect existing project constraints, avoid unnecessary user interruptions, and maintain a persistent, verifiable record of decisions across different sessions.

Highlights

  • Automated Probing: Automatically searches your repository, git history, and documentation to answer questions instead of bothering the user.
  • Persistent Contracts: Saves every intent as a .intent.yaml file, allowing future agents or teammates to start from your established decisions rather than from scratch.
  • Safety Gates: Refuses to execute if a request is ambiguous or touches an irreversible boundary, preventing the agent from making dangerous "guesses."
  • Verification: Checks the final delivered work against the original IntentSpec to ensure the agent actually followed the agreed-upon constraints.
  • Zero-Install: Designed as a portable skill that can be added to existing agent workflows without complex infrastructure.

Related

  • Project
    ZhangHanDong/agent-spec

    An intent compiler for AI agent coding that transforms human requirements into verifiable Task Contracts to ensure AI-generated code is mechanically validated.

  • Dispatch
    Beyond Smarter Agents: Using Formal Verification Gates for AI Coding Loops

    Explore how shifting from prompt-based constraints to compiler-enforced type systems and formal verification creates 'structural backpressure' that forces AI agents to produce correct code.

  • Project
    intent-hq/intent

    Intent is a desktop app that coordinates AI coding agents, notes, tasks, git and terminals via a Rust backend daemon (`intentd`). All state lives locally in SQLite, accessed through JSON‑RPC over a Unix socket or secure WebSocket. The packaged app bundles the daemon; advanced users can install the daemon separately. It supports multiple AI providers, optional GitHub/Linear/Sentry integrations, and an upcoming iOS companion. No telemetry is sent; updates are pulled from GitHub releases only.

  • Project
    NVIDIA-AI-Blueprints/llm-router

    An experimental multimodal router that analyzes user prompts to recommend the optimal LLM or VLM based on trade-offs between accuracy, speed, and cost.

  • Project
    microsoft/agentrc

    A context engineering tool for AI coding agents that automatically generates, measures, and maintains the instruction files and configurations needed for agents to understand a codebase.