Narria: an open-source orchestrator for AI agent workflows

Observe token usage. Control agent budgets. Route between providers. Understand where your tokens go in multi-agent systems.

Open Source Prototype Infrastructure Tool

A tool for developers building AI workflows with multiple agents, each with independent providers and budgets.

What Narria is

Narria is an orchestration layer that sits between your workflow and AI providers. It manages agent lifecycle, tracks token usage, enforces budgets, and routes requests to different providers. It's designed to be transparent, inspectable, and owned by the developer.

Orchestration layer

Manages multiple agents, coordinates their execution, tracks state, handles task dependencies, and returns structured results.

Token tracking

Records every API call: input tokens, output tokens, cost. Attributes tokens to specific tasks, steps, agents, and operations.

Budget enforcement

Sets and enforces spending limits at workflow, task, and agent levels. Stops execution if limits are reached.

Why this exists

Multi-agent AI systems have fundamental observability and control challenges that existing tools don't address.

No per-agent isolation

Most frameworks run all agents through the same provider and model. There's no way to route different tasks to different providers or set per-agent budgets.

Token usage is opaque

You know the total cost after execution. But where did tokens actually go? Which agent spent the most? Which operation is expensive?

Budget control is global-only

You can set a total workflow budget, but not limits per agent or task. One expensive agent can consume the entire budget.

No attribution system

When a task costs 100 tokens, you don't know the breakdown: how much on context, how much on retries, how much on tool calls.

Architecture and components

Workflow schema

Define agents, roles, and dependencies. Specify which provider and model each agent should use, and set individual agent budgets.

Provider adapters

Abstraction layer for different LLM providers. Currently: OpenRouter. In progress: OpenAI, Anthropic. Easy to extend.

Task executor

Runs tasks, handles provider calls, tracks responses. Supports retries, fallbacks, and timeout handling.

Budget engine

Calculates token costs in real-time. Enforces spending limits. Stops tasks if budget is exceeded. Supports granular limits.

Trace system

Records execution traces. Captures all API calls, responses, token counts, latency, and errors for every step.

Analytics engine

Processes traces to generate reports. Breaks down token usage by task, agent, operation, and provider.

Token analytics and attribution

Narria processes execution traces to understand token distribution. Every trace records: API calls, tokens used, latency, errors, and which agent/task made the call.

Global statistics

Total Tokens
245,632
Input Tokens
156,240
Output Tokens
89,392
Estimated Cost
$3.42

Per-task statistics

Task Provider Tokens Cost
Code analysis Anthropic 98,240 $1.47
Code generation OpenAI 87,120 $1.23
Code review OpenRouter 60,272 $0.72

Cost breakdown by operation type

Context Loading
42%
Task Execution
35%
Tool Calls
15%
Retries & Overhead
8%
"The goal isn't to reduce costs. The goal is to understand exactly where tokens are being spent."

Provider architecture

Narria abstracts away provider-specific differences. Each agent is configured with a provider and model independently.

OpenRouter ✓ OpenAI – in progress Anthropic – in progress More providers planned

How it works

Workflow definition

You define a workflow with agents. Each agent has a role, a target provider, a model, and a budget.

Task execution

Narria executes tasks by routing them to the correct agent and provider. It tracks every API call and response.

Token accounting

Every response is parsed. Input tokens, output tokens, and cost are recorded. Budgets are checked in real-time.

Trace analytics

After execution, traces are analyzed. You get detailed reports of where tokens were spent and why.

Open-source architecture

Narria is built in the open. The code is public, the design is transparent, and the development is community-driven. Anyone can review, contribute, or fork.

View Repository Open Issues

Roadmap

Current direction

Project concept defined
Open-source repository created
Landing page and positioning in progress
Workflow orchestration model being designed
Cost attribution model being defined

Planned next steps

Workflow schema
Provider adapters
Task execution tracing
Budget enforcement engine
Token analytics reporting
Initial CLI experience

FAQ

Is Narria production-ready? +
No. Narria is currently an open-source prototype-stage project. It's being developed as a foundation for AI workflow orchestration and cost observability, but production deployments should be evaluated carefully.
Is Narria open source? +
Yes, Narria is being developed as an open-source project on GitHub. The code, design decisions, and roadmap are public and available for contribution.
Who is it for? +
Developers and small teams building AI-assisted development workflows. Anyone who wants to understand and control token spend in multi-agent systems.
Does Narria replace coding assistants? +
No. Narria adds orchestration, budgeting, and observability around AI workflows. It works alongside existing coding assistants and AI frameworks.
Why per-agent providers? +
Different workflow steps have different cost, latency, and quality needs. A planner might need raw reasoning power, while a utility function might need speed. Per-agent provider configuration gives teams flexibility to optimize each role independently.
How can I contribute or get in touch? +
You can visit the GitHub repository to open issues, submit pull requests, or start discussions. You can also contact the maintainer directly by email for feedback, collaboration, or API credit inquiries.

Get involved

Narria is an early-stage project. If you're working on AI agent orchestration, cost observability, or have infrastructure needs, reach out.

sooosooor49@gmail.com