AI Tool Watch

License: MIT Latest release

AI Tool Watch

Plain-English reference for AI capabilities, plans, constraints, and implementations — for humans and agents. AI Tool Watch

Who this is for

Anyone — humans or agents — who needs current, plain-English answers about AI tool capabilities, plan gates, and constraints before committing to a tool.

What problem it solves

AI tool capabilities, plan limits, and constraints change constantly and are scattered across marketing pages. AI Tool Watch is a single, verified, plain-English reference for humans and agents.

Canonical URL

https://aitool.watch/

What is this?

A single source of truth for answering questions like:

Built for AI facilitators, educators, developers, and AI agents that need accurate, current information about AI tool availability. Browse the site, query the JSON API, or connect an agent via MCP.

How it's built

There is no database, no framework, and no dependencies. Every piece of data lives in plain markdown and YAML files under data/. A single zero-dependency Node.js script (scripts/build.js) reads those files and generates the entire site — HTML pages, JSON API, bridge pages, sitemap — into docs/.

That's the whole stack: markdown files, one build script, and Git.

Contributing doesn't require a dev environment. You edit a .md or .yml file, open a PR, and CI rebuilds the site. If you can read a markdown table, you can read (and fix) the data. See design/ARCHITECTURE_PATTERNS.md for the full architectural rationale.

What's covered

18 capabilities, 87 implementations, and 9 open-model access records across major consumer AI products and self-hosted runtimes.

All prices are in USD. Availability reflects the United States region by default.

ProductProviderType
ChatGPTOpenAIHosted
ClaudeAnthropicHosted
CopilotMicrosoftHosted
GeminiGoogleHosted
PerplexityPerplexity AIHosted
GrokxAIHosted
Meta (Llama)MetaOpen model
MistralMistralOpen model
DeepSeekDeepSeekOpen model
Alibaba (Qwen)AlibabaOpen model
OllamaOllamaRuntime
LM StudioLM StudioRuntime

Scope criteria and watchlist: design/SCOPE.md, design/WATCHLIST.md.

Features

Site

API and agents

Automation

Accessibility

This site is designed to meet WCAG 2.1 AA standards:

See also: skill-a11y-audit, a companion project that automates WCAG audits as a reusable AI skill.

Analytics and privacy

Production pages use PostHog for cookieless site-use measurement in the shared Snap Synapse analytics project. Events are accepted only from aitool.watch; persistence is memory-only, person profiles require explicit identification, and session replay masks all page text and form inputs. The static site has no accounts or server-side user store. Direct JSON API reads and local MCP use are not measured by PostHog.

See SECURITY.md for the exact client-side configuration and data boundary.

How data stays current

The verification workflow checks the due inventory on schedule, but paid provider calls remain blocked unless scheduled verification is explicitly approved. Separately approved manual runs can query Gemini, Perplexity, Grok, and Claude within repository spend caps. Models are skipped when verifying their own vendor's products. A change is only flagged when enough independent evidence agrees. Nothing is auto-merged; confirmed changes are surfaced as GitHub issues for human review.

Link integrity is checked every Saturday. Features carry a Checked date; anything older than 7 days is prioritized in the next run.

See VERIFICATION.md for full documentation, commands, and API key setup.

Contributing

Found outdated info? Want to add a feature? See CONTRIBUTING.md.

Quick version:

1. Edit the relevant record in data/platforms/, data/model-access/, data/products/, or data/implementations/

2. Include or preserve the evidence source link

3. Run node scripts/sync-evidence.js

4. Run node scripts/validate-ontology.js

5. Submit a PR

Local development

git clone https://github.com/snapsynapse/ai-tool-watch.git

cd ai-tool-watch



node scripts/build.js               # Build the site


node scripts/sync-evidence.js       # Sync evidence records


node scripts/validate-ontology.js   # Validate cross-record integrity



open docs/index.html

Build output:

Data format: data/_schema.md. Ontology schema: design/SCHEMA_PROPOSAL.md.

Deployment

The site prepares and deploys one reviewed artifact on every push to main and on its Monday/Thursday scheduled run. Pull requests validate the source tree and prepare the artifact but do not deploy it.

How it works

1. Validate and prepare jobs (.github/workflows/build.yml)

- Validates the committed source/evidence tree, then runs node scripts/prepare-publication.js

- Synchronizes canonical evidence, generates docs/, validates the generated site and a human/API/MCP fixture, and writes docs/publication-manifest.json

- If the scheduled or push run changes canonical inputs or generated output, commits the scoped source and docs/ paths with [skip ci]

- Uploads the resulting reviewed-publication artifact, including hidden files

2. Deploy job (same workflow)

- Downloads and verifies the reviewed artifact in a fresh directory before uploading it to GitHub Pages

- Runs on pushes to main and scheduled builds, not PRs

3. FTP deploy (.github/workflows/deploy-ftp.yml)

- Reuses the same reviewed artifact, verifies it in a fresh directory, then deploys it to https://aitool.watch/ via FTP

- Runs in parallel with the GitHub Pages deployment on push-to-main and scheduled runs

- Requires FTP_HOST, FTP_USER, FTP_PASS secrets

The artifact boundary is local and CI evidence only. It does not establish a hosted deployment, direct-email delivery, or independent missed-run monitoring.

GitHub Pages setup

To enable GitHub Pages on a fork:

1. Go to Settings → Pages

2. Under "Build and deployment", select GitHub Actions

3. The workflow will deploy to https://.github.io/ai-tool-watch/

Design documentation

Architecture, ontology, and project status docs live in design/:

Skills: skills/ (README)

Sponsor

AI Tool Watch is free and open. If you rely on this reference, consider sponsoring its upkeep to keep the verification pipeline running. See SPONSORS.md.

Stewardship and use

AI Tool Watch is an open reference under Snap Synapse LLC stewardship, authored by Sam Rogers. The reference is MIT-licensed and free for any use. Purpose, invariants, and the current review decision are in INTENT.md.

License

MIT - see LICENSE

Credits

Created for educators, writers, builders, and the larger AI community.

With help from Claude Code, of course.

Found an error? Open an issue or submit a PR!