Letter
A coding agent for growth engineering that measures how AI answers name and cite you, then fixes and builds your site through pull requests.
What it does
Letter (from The Letter Company) asks AI answer engines the questions you care about and records how often the answers name or cite your brand against competitors (share of voice). Its Visitors script and server snippet show the AI agents and crawlers reading your site and the visits that arrive from AI answers. Letterpress-1 then turns what it finds into work: it fixes issues on your marketing site, builds pages such as comparisons and answers, and opens each change as a pull request with a live preview, so nothing ships until you merge it. Everything in the app is also exposed through an API, a CLI and an MCP server, so Claude Code, Codex or your own agent can run the same tools.
Who it suits
Developers, founders and agent-driven teams who want API and MCP access and want the fixes shipped as pull requests, not just reported.
Who should skip it
You want a no-code dashboard for a marketing team to watch share of voice, with no one to review pull requests.
Strengths
- API, CLI and MCP on every plan; your own agent gets the same tools as the app
- Acts on findings: site fixes and new pages arrive as pull requests with previews
- Sees both sides: AI answer probes plus AI crawler visits and click-throughs on your own site
- Usage pricing published to the token, from $25 a month, with free tokens to start
Limits
- Built around a git repository; teams without one get less from it
- Its list of answer engines probed is not on its pricing page
- Less of a point-and-click reporting dashboard than Peec AI or Otterly.AI
- Cost
- From $25/month (Hobbyist, 2,500 tokens) to $2,500/month (Scale); Enterprise $10,000/month. Each AI visibility answer read costs 12 tokens. Every account starts with 1,000 free tokens, no card.
- Engines
- Not listed on the pricing page. Check the API docs at app.letterstory.com/docs/api or ask Letter which engines its probes cover.
- API and MCP
- API, CLI and MCP on every plan, able to read data and start changes (pull requests).
- Sources
- Pricing · Product overview · API docs · agents.md
Criterion scores
- Citation tracking coverage 25%9/10Tracks named and cited rates and share of voice on its own questions, plus AI crawler visits and click-throughs from AI answers on your site.
- Answer engines covered 15%4/10The engines its probes cover are not published on the pricing page, so it scores as an unpublished list.
- API and MCP access 15%10/10API, CLI and MCP are on every plan and can start work, not only read it.
- Pricing transparency 10%10/10Every plan, per-token rate and overage is published; self-serve start with free tokens.
- Actionability 20%10/10Opens pull requests that fix the site and build pages, each with a preview for review.
- Evidence quality 15%7/10Share of voice is computed from the probe answers it reads; check in a trial how much of each raw answer and its sources you can open.