CLI dashboard · real-time · local

Full observability
into your AI agents.
As it happens.

AgentWatch gives developers a real-time CLI dashboard showing every API call, every token spent, and every agent action as it happens. Zero guesswork. Zero surprises. Zero cost overruns.

macOS 12 Ventura or later · Python 3.10+ · No account

agentwatch v1.2.0 — live
ACTIVE
agent: gpt-4o-2024-11-20 framework: LangChain
updating live
<100ms
Dashboard refresh rate
12+
AI agent frameworks supported
0 bytes
Sent to any external server
Capabilities

Everything you need
to own your agent costs.

Built for developers who want full control, not just dashboards.

Real-time cost tracking

Watch costs accumulate as they happen. Per-call, per-session, and per-agent breakdowns updated at 100ms intervals.

Full API call history

Every call logged with timestamp, tool name, token counts, latency, and cost. Searchable and filterable. Goes back to session start.

Multi-agent monitoring

Track multiple agents running simultaneously. Compare cost rates, call frequencies, and behavior patterns side by side.

Behavior visualization

Spot loops, redundant calls, and inefficiencies in agent decision-making before they compound into runaway costs.

Budget alerts

Set per-agent or per-session spend limits. Get notified — or automatically halt agents — before costs spiral out of control.

Export & analyze

Export full logs as JSON or CSV. Pipe into your own analytics tooling, Jupyter notebooks, or billing systems.

Supported frameworks

Works with your stack.

Drop-in integration for the most popular agent frameworks.

LangChain
AutoGen
CrewAI
LlamaIndex
DSPy
Claude Code
Agno
Custom SDK

Any framework that uses OpenAI-compatible APIs works out of the box.

100% local

Your agent's actions are your business. Nobody else's.

AgentWatch runs entirely as a local CLI process. All telemetry, logs, and cost data stays on your machine. We have no servers, no analytics pipeline, and no visibility into your agents.

No outbound network connections
Logs stored only in local filesystem
Open source — audit every line
Air-gapped environments supported
agentwatch.config.ts local only
// All data stays on your machine
export const config = {
storage: '~/.agentwatch/logs',
telemetry: false,
network: false,
refreshMs: 100,
}
Get started in one command
Install the CLI, wrap your agent, and you're monitoring in under 60 seconds.
$ pip install agentwatch

The first time I ran an AI agent overnight I woke up to a $47 bill and no clear idea what it had actually done. The agent had finished its task — sort of — but somewhere in the middle it had looped on a subtask, made the same API call sixteen times, and burned through tokens I hadn't budgeted for. I had no log, no trace, no replay. Just a number on a dashboard and a vague sense of unease.

I looked around for something that would show me, in real time, what my agents were doing. The tools I found were all cloud-based: pipe your agent's traffic through our servers, sign up for an account, and we'll show you a pretty dashboard. That meant handing off my prompts, my responses, and my architecture to a third party whose business model I didn't fully understand. For production agents working with sensitive data, that felt like exactly the wrong trade.

AgentWatch is what I actually wanted: a local CLI process that intercepts nothing, sits beside your agent as it runs, and gives you full cost and behavior visibility without a single byte leaving your machine. It's not a SaaS. It's a tool — yours, running locally, answering only to you.

If this tool is useful, consider supporting its development.

Buy me a coffee
Quick start

Full observability in four steps.

01

Install the CLI

Run pip install agentwatch or clone the repo. No account, no signup — it's ready to use the moment it's installed.

02

Point it at your agent

Add one line to your agentwatch.config.ts to point at your framework — LangChain, CrewAI, AutoGen, or any OpenAI-compatible SDK. No code changes to your agent.

03

Watch costs and behavior in real time

Run agentwatch --start and your terminal becomes a live dashboard — every API call, token count, latency, and running cost updated at 100ms intervals.

04

Everything stays local — zero cloud

Logs are written to ~/.agentwatch/ and never leave your machine. Export as JSON or CSV for your own analysis. Full observability, no third-party servers, no exceptions.

$ agentwatch --start

Stop guessing.
Start watching.

Full observability for your AI agents. Free, local, and open source.

macOS 12+ · Python 3.10+ · No account · Open source · MIT License