Mass AI Atlas

The living map of Massachusetts' AI ecosystem — built by the community for the community and founders, presented by The Open Accelerator (an initiative between the MA AI Hub, Red Hat & IBM).

Four ways to explore the Massachusetts AI ecosystem:

• 🗺️ Map — every VC, lab, accelerator and community pinned across Massachusetts
• ✨ Galaxy — an animated network of how the ecosystem connects (spinouts, funds, hosts)
• 🧭 Founder Journey — resources curated by stage, from first spark to scale
• 📇 Directory — full-text search with category and stage filters

Each entry's drawer says what the organization offers founders, shows when a maintainer last verified it, and under places to meet lists the cafés, quick bites and quieter spots within a five-minute walk. Every entry and view is linkable — ?entry=csail opens straight to an organization, #galaxy to a view (there's a 🔗 Copy link button in each entry's drawer).

Building on this as a template? Start with HACKATHON.md: setup, the architecture on one page, where to extend it, and what to change because it is still pointed at The Open Accelerator.

Run it

It's a static site — no build step.

The hosted live page is at https://the-open-accelerator.com/ecosystem/

# locally
python3 -m http.server     # then open http://localhost:8000

Opening index.html straight from file:// doesn't work: the page fetches its data from data/*.yml, which browsers block on file://. Use the local server above.

The maintenance scripts (scripts/) need Node 20+ and the two npm dependencies (js-yaml, Ajv): run npm ci once.

Deploy

Merging is publishing. A GitHub Action deploys index.html, atlas-data.js, data/ and the map style to the live site on every merge to main that touches them — it re-validates the data first, then rsyncs to the server and posts a summary to the TOA Slack. (Not GitHub Pages.)

Manual fallback, from a checkout with SSH access to the server:

./scripts/sync-to-site.sh <destination>

The page must be served at a directory URL (…/ecosystem/, trailing slash) so the relative data/ fetches resolve.

If the repo ever moves (or you host your own copy): set GH_REPO in index.html (search for const GH_REPO) to your org/repo so the in-app Suggest and Flag buttons point at the right issue forms.

Contribute

The atlas is community-maintained:

• ➕ Suggest an entry — open a suggestion or edit data/entities.yml and send a PR
• 🚩 Flag an entry — report something outdated or miscategorized (or use the Flag button on any entry inside the atlas)

See CONTRIBUTING.md for the entry schema and review process. Every PR is automatically validated by CI, and a maintainer can hand your suggestion to the atlas bot, which drafts it as a pull request (geocoded and validated) for human review — nothing publishes without a maintainer merge.

Maintainers turn accepted issues and PRs into entries — see MAINTAINING.md for the step-by-step process.

Repo layout

index.html                        the app — one file of HTML/CSS/vanilla JS (Leaflet +
                                  MapLibre GL, Tailwind & web fonts via CDN; TOA-branded)
data/entities.yml                 the dataset — this is what you edit
data/categories.yml, stages.yml   the categories and founder stages
data/*.schema.yml                 JSON Schemas for the three data files
atlas-data.js                     shared loader: parses data/*.yml + validates against
                                  the schemas (used by index.html and scripts/)
package.json                      npm deps for the scripts (js-yaml, Ajv) — `npm ci`
map-style-dark.json               the dark basemap style (OpenFreeMap fiord + our grafts)
scripts/validate.js               schema + integrity validation (runs in CI and locally)
scripts/load-data.js              Node wrapper around atlas-data.js — scripts load data via this
test/data-refactor-equivalence.test.js  one-time check that data/ matches the old data.js
scripts/stale.js                  freshness report — entries longest unverified
scripts/export-csv.js             export the dataset to a spreadsheet-friendly CSV
scripts/issue-to-entry.js         the suggestion bot's form parser / entry builder
scripts/fetch-events.js           events aggregator — feeds /ecosystem/events/ weekly
scripts/build-events-page.js      assembles events/index.html (generated — see docs/)
scripts/trigger-weekly-events.sh  server-side cron that starts the weekly refresh (GitHub cron is late)
scripts/sync-to-site.sh           manual deploy fallback (CI normally deploys on merge)
.github/ISSUE_TEMPLATE/           suggest & flag forms
.github/workflows/validate.yml    CI validation + freshness summary on every data PR
.github/workflows/deploy.yml      auto-deploy to the live site + Slack notification
.github/workflows/suggest-to-pr.yml  the suggestion bot (maintainer-gated, drafts PRs)
.github/workflows/weekly-events.yml  Monday 7am ET events refresh + Slack post
docs/events-guide.md              how the events page & weekly refresh work
CONTRIBUTING.md                   entry schema & contributor workflow
MAINTAINING.md                    maintainer guide — turn issues & PRs into data/entities.yml
PROJECT_STATE.md                  architecture, decisions & backlog (start here if new)

The data

The atlas is also an open dataset — the only maintained, human-verified dataset of Massachusetts' AI ecosystem. Two machine-readable endpoints regenerate on every deploy:

• JSON — https://the-open-accelerator.com/ecosystem/atlas.json — a versioned envelope (schema: 1) with all entries, categories, stages and per-category counts
• CSV — https://the-open-accelerator.com/ecosystem/atlas.csv — one row per organization, ready for Google Sheets (File → Import) or Excel (Data → From Text/CSV)

Columns/fields include name, category, location, lat/lng, website, description, tags, founder stages and the month a maintainer last verified the entry.

To regenerate locally: node scripts/export-json.js and node scripts/export-csv.js.

Methodology: every entry is checked by a human against the organization's own website plus an independent source before it ships, and re-verified on a rolling basis (lastVerified records the month; entries more than a year unverified are flagged in the app). Defunct or unverifiable organizations are removed.

License: the dataset is licensed under CC BY 4.0 — free to use, share and adapt with attribution to The Open Accelerator. Suggested citation: "Mass AI Atlas, The Open Accelerator — the-open-accelerator.com/ecosystem".

Some pin locations are approximate (neighborhood-level) and community groups move between venues — check official sites before visiting. Curated since June 2026.

© 2026 The Open Accelerator.