Local-first AI observability

See every AI token you spend.

Free early access for technical testers on macOS Apple Silicon. One local-first hub for agent setups, local models, and multiple cloud APIs.

Get Started → View on GitHub

Free early access
macOS Apple Silicon
Local-first usage hub
Browser dashboard

Your AI Tools

Cursor Claude Codex Any Tool

TokenPulse Proxy

⚡ Token counting 💰 Cost tracking 💾 Local SQLite logging

AI Providers

OpenAI Anthropic Google Mistral Groq OpenRouter Ollama

Local Dashboard

📊 Spend & trends 🎯 Budgets & alerts 🔍 Optimizer signals

The bottleneck

The problem with AI spending

The more models and providers you use, the harder it gets to understand where cost is coming from and what to change.

Scattered billing

Juggling 4+ provider dashboards to understand what you're spending.

Invisible local usage

Running Ollama and LM Studio with zero visibility into what they're doing.

No single view

No way to compare cloud costs vs local performance in one place.

How it works

Three steps. Five minutes.

TokenPulse sits in the request path, records the useful parts, and gives you one local dashboard instead of eight tabs.

Point your tools at TokenPulse

Change one environment variable. That's it.

export OPENAI_BASE_URL=http://localhost:4100

TokenPulse intercepts and logs

Every request flows through the proxy. Usage metadata is stored locally in SQLite.

Open your dashboard

The main dashboard runs in your browser at http://127.0.0.1:4200.

Dashboard

Your AI command center

Track live requests, budget pressure, model performance, and cost-saving opportunities from one dashboard built for daily use.

TokenPulse dashboard overview with live stats and request feed

TokenPulse dashboard charts and provider breakdowns

TokenPulse budget tracking and threshold controls

Features

Built for power users

Everything important stays local, but the visibility is good enough for serious day-to-day optimization work.

🔌

10+ Providers

OpenAI, Anthropic, Google, Mistral, Groq, Ollama, LM Studio, OpenRouter — all through one proxy.

💰

Real-time cost tracking

Per-request token counts and cost calculation as requests flow through.

📈

Budget alerts

Set spending limits, get notified before you blow through them.

🏠

Local-first

SQLite on your machine. No cloud account, no telemetry, no data leaving your network.

⚡

Streaming support

Full SSE streaming passthrough with live token counting.

🔍

Cost optimizer

Detects model misuse and suggests cheaper alternatives that fit.

📊

Forecasting

Monthly spend projections based on your actual usage patterns.

📤

CSV export

Export everything for spreadsheets, reports, or your own analysis.

Comparison

How TokenPulse compares

Local model visibility and zero-account setup are the main differentiators, especially if you already run AI tools on your own machine.

Feature	TokenPulse	Helicone	LiteLLM	CostGoat
Local model tracking	✅	❌	Partial	❌
Real-time proxy	✅	✅	✅	❌ (polls)
100% self-hosted	✅	✅ (OSS)	✅ (OSS)	❌
No account for core features	✅	❌	❌	❌
Data stays local	✅	❌	❌	✅
Price	Free (early access)	Free-$500/mo	Free	$9/mo

Getting started

One command install for v1 early access

TokenPulse v1 is free early access for technical testers on macOS Apple Silicon. Install it, then open the browser dashboard.

 macOS Apple Silicon 🧪 Early access testers

Linux, Windows, and NVIDIA setups are planned for later releases.

Read the full getting started guide

curl -fsSL https://raw.githubusercontent.com/TokenPulse26/TokenPulse/main/install.sh | bash

# Proxy
http://127.0.0.1:4100

# Main browser dashboard
http://127.0.0.1:4200

Agent-installable

Set up agents quickly

Follow the agent guide to connect your coding agents and tester workflows to TokenPulse with a local-first setup.

Open AGENT_SETUP.md

OpenClaw ready

Use TokenPulse with OpenClaw

TokenPulse includes an OpenClaw setup path for early testers. Start with the official setup guide.

Open OPENCLAW_SETUP.md

Verify your install

Run the built-in tester check

TokenPulse now includes ~/.tokenpulse/agent_verify.py for agent and tester verification, plus a first-tester QA pack.

python3 ~/.tokenpulse/agent_verify.py

Route examples

Use the right local endpoint

These are common base URLs for v1:

OpenAI-compatible: http://localhost:4100
Anthropic:        http://localhost:4100/anthropic
Ollama:           http://localhost:4100/ollama
LM Studio:        http://localhost:4100/lmstudio