Show HN: Self-hosted monitoring for AI recommendations (MIT License) Lettertrace, a new open-source tool released under the MIT License, lets companies self-host monitoring of how often AI assistants like Claude, ChatGPT, and Gemini mention their brand. The developer-first tool, which works with Anthropic, OpenAI, and Google keys, provides metrics such as visibility, share of voice, prominence, and sentiment, with a sample run showing Acme mentioned in 15 of 24 prompts (41% share of voice) and a positive sentiment score of +0.34. Lettertrace measures how often Claude, ChatGPT, and Gemini mention your company. But there's a catch: it's free end-to-end, developer-first, and open source. works with ChatGPT, Claude & Gemini · self-host in minutes $ lettertrace run --project acme › querying claude-opus-4-8 across 24 prompts… ✓ acme mentioned in 15 / 24 answers share of voice ████████░░░░░░░░ 41% sentiment +0.34 positive top competitor notion · 28% $ how it works Add the subjects your buyers ask about, “best CRM for startups”, “email tools for agencies”. Lettertrace writes the real questions people put to AI assistants, so you monitor how they’re actually answered. Every run is a datapoint. Watch visibility, prominence, and sentiment move week over week. Ingest rival brands and see your share of voice, who the models recommend, and how often. the numbers that matter Lettertrace turns raw AI answers into the metrics AEO/GEO teams track: visibility, share of voice, prominence, and sentiment, per topic, per model, over time. 62% 41% +0.34 5 sample answer · claude “For fast-growing startups, a few tools stand out. Acme is a strong pick for teams that want automation without losing their brand voice, while Notion and Linear are popular for docs and issue tracking…” features Use your own Anthropic, OpenAI & Google keys. They’re encrypted at rest and never leave your infrastructure. Monitor Claude, ChatGPT, Gemini, and Google AI Overviews side by side. Add more answer engines as they matter. Turn one topic into dozens of natural prompts automatically, no manual prompt-writing. Know not just if you appear, but whether the answer speaks well of you, and recommends you. See exactly how often you win the answer versus each competitor you track. Run daily or weekly on autopilot and build a trend line you can act on. open source $ git clone lettertrace && cd lettertrace $ npm install $ cp .env.example .env.local add Supabase URL + keys $ npm run dev → http://localhost:3000 Set up your brand, add a key, and run your first monitor in minutes.