Best AI visibility tools for open-source companies (2026)
Coding agents now pick libraries for developers. How open-source companies can track Claude Code, Codex and OpenCode recommendations, docs crawling and Reddit citations.
An open-source company has two funnels, and AI now sits at the top of both. The first is adoption: a developer needs a job queue, a vector store or an auth library, and asks a coding agent to add one. The agent picks a package and writes the code. The second is revenue: a team already running your project asks ChatGPT whether managed hosting is worth it, or which vendors offer enterprise support.
Most AI visibility tools only see the second funnel. For open-source companies, the first one is where the long-term numbers come from.
How open-source projects get chosen now
A developer working in Claude Code, Codex or OpenCode rarely types "what's the best queue library". They type a task: "add background jobs to this service". The agent decides which library fits, often based on what it has seen most in training data and docs, and sometimes after searching the web. If your project isn't the default, you lose installs you never knew about.
The commercial questions look more like classic B2B prompts. "Managed [project] hosting", "[project] enterprise support", "[project] vs [proprietary product] total cost". These are answered from your docs, comparison posts, GitHub discussions and Reddit threads.
Prompts worth tracking
- In a coding agent: "Add a job queue to this Node.js project"
- In a coding agent: "Set up full-text search for this Postgres app"
- "Best open-source alternative to [proprietary tool]"
- "[Your project] vs [competing project] performance"
- "Is [your project] production ready?"
- "Managed hosting for [your project]"
- "Can I use [your project] commercially under its license?"
The license question is worth watching closely if you've changed licenses. Engines can describe the old terms for a long time.
What a tool needs for open source
Coding agent visibility
Promptwatch tracks what Codex, Claude Code and OpenCode recommend: tools, libraries and docs. That's the part of the adoption funnel nobody could measure before. We run task-style prompts across a few stacks (Node, Python, Go, whatever your users run) and compare how often the agent reaches for your package versus a competitor's.
Expect the three agents to disagree. OpenCode, for one, searches the web through Exa. When one agent picks your package and another never does, that gap is worth digging into. Read the answers side by side, then check whether AI crawlers are reaching your docs at all.
Docs crawling
Your documentation is your most cited asset, if crawlers can read it. Agent Analytics, from Professional ($245/mo), records AI crawler visits verified against operator IP ranges: GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot and others. A per-page citation rate tells you whether a docs page is crawled and ignored, or never crawled at all. Crawler logs can come in through Cloudflare, Vercel and Netlify log drains, Fastly and other CDNs, which suits docs sites hosted on those platforms.
One finding from Promptwatch's data reports is relevant here: Markdown is 0.05% of AI search citations. Raw README and .md files in a repo are rarely what gets cited. Rendered HTML docs pages are.
GitHub and Reddit citations
Developer answers cite GitHub repositories, issues and discussions, which show up as domains and URLs in Promptwatch's citation analytics, with competitor domains highlighted. Reddit has its own report with subreddits, upvotes and comments. Be careful with Reddit plans, though. Promptwatch's data showed Reddit's share of ChatGPT citations falling from about 4% to about 0.5% on August 14, 2026. Check your own answers before building a community strategy around it.
API and MCP access
Developer-run companies like to script things. Promptwatch's MCP server and REST API are on every paid plan, so your team can pull visibility data into its own tooling or ask Claude or Cursor about it directly.
The ranking
1. Promptwatch
The platform we use for open-source clients. Coding agent visibility covers the adoption funnel; prompt tracking, citation analytics, crawler logs and visitor analytics cover the commercial one, all in the same workspace. Visitor analytics counts AI-referred visits and conversions, so cloud signups from ChatGPT or Perplexity are visible. Essential is $95/mo, Professional $245/mo with crawler logs, Business $579/mo, Enterprise custom. Every self-serve plan has a 7-day free trial on monthly billing.
2. AthenaHQ
A GEO platform with a score, an Action Center and a content agent. Starter is $295/mo for 3,600 credits across 11 models. API access is a paid add-on on Starter. Our catalog lists no coding agent tracking or crawler logs.
3. The Prompting Company
Made for developer tools, APIs and infrastructure, and the closest specialist for this use case. It tracks ChatGPT, Perplexity and Gemini from $99/mo, and it can test whether coding agents complete a real onboarding task with your docs. Coding agent tracking (Claude Code, Codex, OpenCode) is on White-glove at $3,000/mo.
4. Scrunch AI
Core is $250/mo for 125 prompts on ChatGPT, Perplexity, Google AI Overviews and Copilot, with bot traffic reporting. Fine for monitoring the commercial prompts if coding agents aren't a concern.
5. Lettertrace
A free, MIT-licensed monitor you can self-host or use hosted, running prompts with your own API keys against Claude, ChatGPT, Gemini, Google AI Overviews or Perplexity Sonar. It has a REST API, an MCP server and a CLI. Good for an engineering team that wants something inspectable and cheap, and doesn't need crawler logs or traffic attribution.
6. Ansvisor
An open-source AI search product under the MIT license. Self-hosting needs Supabase, an AI provider and a scraping service; Ansvisor Cloud starts at $49/mo for ChatGPT and Perplexity. Our catalog notes that its own pages disagree on trial length and seat counts, so read the current terms.
7. GetCito
Offers a free, MIT-licensed self-host tracker for ChatGPT and Google AI Overviews on a weekly refresh, or managed plans from $599/mo with a human team writing content. The self-host option is a light first look.
Comparison
| Tool | Starting price | Coding agents | Crawler logs | Self-host option |
|---|---|---|---|---|
| Promptwatch | $95/mo | Codex, Claude Code, OpenCode | From Professional ($245) | No |
| AthenaHQ | $295/mo | Not listed | Not listed | No |
| The Prompting Company | $99/mo | From $3,000/mo | Not listed | No |
| Scrunch AI | $250/mo | Not listed | Bot traffic reporting | No |
| Lettertrace | Free (you pay model APIs) | Not listed | Not listed | Yes (MIT) |
| Ansvisor | $49/mo cloud | Not listed | Not listed | Yes (MIT) |
| GetCito | Free self-host; managed $599/mo | Not listed | Not listed | Yes (MIT) |
Who should pick which
If your growth depends on developers choosing your package by default, coding agent visibility is the measurement that matters, and Promptwatch has it next to everything else. The Prompting Company is the alternative for devtools teams that also want onboarding tests run by agents. A small project with no commercial arm can learn a lot from Lettertrace or a self-hosted tracker for the price of API calls. Once you sell hosting or support and need to attribute signups, you'll need crawler logs and visitor analytics, which the free tools don't cover.
Honest limits
Coding agent results depend on the model version, the project context and the exact task. An agent working in a repo that already uses a competitor will keep using it. We read these results as trends over weeks. No tool can make an agent pick your library, and training data changes on the model vendors' schedule, not yours.
FAQ
Can we track what Claude Code recommends?
Yes. Promptwatch tracks Codex, Claude Code and OpenCode recommendations. Use task-style prompts that mirror what your users actually ask an agent to do.
Do llms.txt files help?
Possibly; it's cheap to add one. Promptwatch has a free llms.txt generator and an Agentic Readiness Check that tests robots.txt and llms.txt. We still put rendered HTML docs first.
Is GitHub a citation source?
Often for developer prompts. Citation analytics shows github.com URLs among cited sources, so you can see which repos, issues or discussions engines lean on.
Next step
Run a few coding agent tasks against your own project: start a 7-day free trial of Promptwatch. If you want us to build the prompt set with you, write to hello@visiblyrank.com.