tma1-ai

tma1-ai / tma1

Public

Local-first observability for AI agents. Tokens, cost, latency, and traces — stored locally, queryable with SQL.

tma1.ai agent-observability ai-agents claude-code codex llm-observability
19
0
100% credibility
Found Mar 17, 2026 at 19 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
JavaScript
AI Summary

TMA1 is a local observability tool for AI agents that tracks token usage, costs, latency, errors, and security signals using an embedded database, accessible via a web dashboard.

How It Works

1
👀 Discover TMA1

You hear about a simple tool that watches your AI helpers' activity right on your own computer, without sharing data online.

2
📥 Install with one click

Run a quick command to download and set up everything you need in a private folder on your machine.

3
🤖 Connect your AI agent

Tell your AI agent to share its thoughts and actions with TMA1 using a simple setting change.

4
🖥️ Open the dashboard

Launch your personal viewer at a local web address to see real-time updates from your agent.

5
📊 Spot insights instantly

Watch costs, speeds, conversations, and alerts pop up, helping you understand exactly what's happening.

🛡️ Monitor safely forever

Enjoy full visibility into your AI's behavior, costs, and safety—all stored privately on your computer.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 19 to 19 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is tma1?

TMA1 delivers local-first observability for AI agents, tracking tokens, cost, latency, traces, and errors—all stored locally and queryable with SQL. Install via curl on macOS/Linux/Windows, run `tma1-server`, and configure agents like Claude Code or OpenClaw to send OTLP data to localhost:14318 for an instant dashboard at the same port. No cloud, Docker, or Grafana needed; it's JavaScript-friendly for agent devs monitoring runs locally.

Why is it gaining traction?

It skips vendor lock-in with embedded storage and direct SQL access, plus user-facing wins like cost projections, cache efficiency, p50/p95 latency tables, and anomaly flags for high tokens or errors. Devs dig the one-command setup and conversation replay for tools like TMA1 scheme or agents, beating clunky cloud alternatives for quick local insights on cost and latency.

Who should use this?

Agent builders tweaking Claude Code, Codex, or OpenClaw in terminals; JavaScript scripters instrumenting OTel for local runs. Ideal for solo devs debugging latency spikes or token burn in prototypes, without infra overhead.

Verdict

Promising for local-first GitHub observability (19 stars, 1.0% credibility score), with strong docs and tests, but early-stage—expect tweaks. Grab it now if agent metrics matter; skip for production-scale needs.

(178 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.