Knowledge / Academic / Enterprise SaaS / MCP Ecosystem

70+ sources across three sub-categories. Includes 25+ notable MCP servers and 20+ unified-API platforms.

A) Knowledge / Encyclopedic / Academic / Research

General-purpose knowledge

• Wikipedia REST + MediaWiki Action API — Free, CC BY-SA/GFDL. 5/5 viability. User-Agent header required (no-UA traffic banned). REST and Action APIs have different shapes
• Wikidata SPARQL — Free, CC0. 110M+ items. WDQS performance degraded through 2025-26 — 60s query timeout, partitioning into "main" + "scholarly" graphs. For heavy work use JSON dump or QLever mirror
• DBpedia — Free, CC BY-SA. Largely superseded by Wikidata
• OpenStreetMap (Overpass + Nominatim) — Free (ODbL). Overpass: 180s default timeout, 512 MiB RAM. Nominatim: strict 1 req/s, must self-host for production
• Wolfram Alpha API — Free 2k non-commercial calls/mo; commercial from ~$25/mo for 1k calls. Output is HTML+plaintext "pods" — semi-structured

Academic / scholarly (the unified-intelligence backbone)

• arXiv API — Free, 1 req per 3 sec single connection. Aggressive 429 enforcement since Feb 2026. AWS S3 requester-pays mirror for bulk
• bioRxiv / medRxiv API — Free, ~<1 req/s polite. Cursor pagination inconsistent
• PubMed E-utilities (NCBI) — Free; 3/sec anonymous, 10/sec with key. PMC backend migrating Feb 2026 — schema drift expected
• Semantic Scholar (S2) — Free; 5,000 req/5min unauth (shared pool, easy to starve), 1 req/s with key. 200M+ papers + citation graph + embeddings + TLDRs
• CORE.ac.uk — Free tier; commercial on request. 250M+ OA outputs; fulltext when available
• OpenAlex — Free API key (required since Feb 2025); $1/day credit default, Premium scales to millions. 250M+ works, CC0. Bulk monthly snapshot (~330 GB)
• Crossref REST — Free; "polite pool" via mailto. Rate-limit headers revised Dec 2025. 180M DOI records, CC0
• ORCID API — Free; ~24 req/s public. Member API for write
• DataCite REST — Free, CC0. Canonical source for dataset/code DOIs
• ROR (Research Organisation Registry) — Free, CC0. 2000 req/5min. Authority for "what counts as one institution"
• Unpaywall — Free; 100k req/day with email=. Single best "is there free fulltext?" oracle
• Google Scholar (unofficial) — Aggressively blocked; do not depend on this. Use OpenAlex / S2

Patents / IP

• USPTO Open Data Portal — Free. PatentsView is research-grade. ODP Beta sunsets May 29, 2026 — migrate to new ODP
• EPO Open Patent Services (OPS) — Free up to 4 GB/week; paid above. 130M+ patents from EPO/WIPO/national offices
• Google Patents (BigQuery) — BigQuery patents-public-data. Old REST API deprecated; SerpAPI provides unofficial shim

Books / library

• DOAJ — Free; vetted OA journals
• JSTOR (Constellate) — Constellate sunset 2024-25; institutional only now
• HathiTrust — Free metadata + Extracted Features (token counts); fulltext gated to affiliated unis
• Europeana — Free; 10k req/day. European cultural heritage
• Library of Congress — Free, US gov works. 20 req/min recommended
• BnF / Gallica — Free
• British National Bibliography (BNB) — Free, CC0 Linked Open Data. BL cyber incident (2023) ripple effects
• Google Books API — Free 1k req/day default. Deprecation risk
• Open Library — Free, CC0. Community-maintained
• ISBNdb — From $14.95/mo
• WorldCat (OCLC) — Membership-gated; OCLC litigious about scraping

B) Enterprise SaaS Connectors (Internal Side)

Messaging / collaboration

• Slack — Free dev; paid Slack workspace for many endpoints. March 2026: conversations.history/replies for non-Marketplace apps dropped to Tier 1 (1 req/min, 15 objects max) — kneecapped non-Marketplace integrations
• Microsoft Graph (M365) — Teams/Outlook/Drive/SharePoint/Calendar/AD. Free with M365 tenant; Teams export gated to E5. Throttling cannot be raised — use Graph Data Connect for bulk. CASA audit ~$5-20k for restricted scopes
• Google Workspace APIs — Gmail (250 quota units/user/sec), Drive (12k req/min/user), Docs, Sheets, Calendar, Admin SDK. Restricted scopes require annual CASA audit (~$5-20k). DWD setup admin-heavy

Wiki / knowledge / docs

• Notion API — Free with Notion plan; 3 req/s average, burst to 10. API version 2026-04-01 returns standard X-RateLimit-* headers. Block-based content model is fiddly
• Atlassian Confluence Cloud — OAuth 2 (3LO) / API token / Forge / Connect. Moving to points-based + tiered quotas from March 2, 2026. v1 vs v2 split messy
• Atlassian Jira Cloud — Same auth/rate limits. JQL rich but slow on big tenants; webhooks essential

Project management

• Linear API (GraphQL) — Free with Linear; 5,000 req/hr (API key) / 1,500 (OAuth); 250k complexity points/hr. Best-designed PM API in market
• Asana — Free with Asana; 150 req/min free, 1500 paid; cost-aware throttling
• ClickUp, Trello, Monday.com, Airtable, Coda — Per-platform; rate limits documented

CRM

• Salesforce REST + Bulk + Pub/Sub — Per-org via Salesforce SKU. 24h API call allocation by edition (Enterprise ~1k/license + 100k floor). You consume the customer's API budget — be courteous. Bulk 2.0 + CDC via Pub/Sub is modern path; avoid Streaming (CometD) legacy
• HubSpot CRM — Free dev; per-account install limits. 110 req/10s per installed account on 2025.2/2026.03 (Marketplace apps); Pro 650k/day, Enterprise 1M/day
• Pipedrive, Zoho CRM — Per-platform OAuth

Support / customer ops

• Zendesk — Suite Pro 700 req/min, higher on Enterprise; per-endpoint sub-caps
• Intercom — 10k req/min per workspace default
• Front — ~50 req/min default
• Help Scout — 400 req/min per app per inbox
• Freshdesk — Plan-tied (Pro 400 req/min)
• Drift — Salesloft acquisition (2024) makes direction murky
• Crisp, Vitally, Gainsight — Per-platform

Analytics

• Mixpanel Export API — 60 req/hr for raw export endpoint — restrictive. Use data pipelines (S3/BigQuery) on Enterprise for bulk
• Amplitude Export API — Tight on export; warehouse sync recommended
• PostHog API — Self-hostable; avoids API ceilings entirely. Best modern story for product analytics extraction
• Heap — Auto-capture hard to "export" cleanly; usually consumed via Snowflake share

Marketing / messaging

• Segment, RudderStack (open-source self-host), Customer.io, Iterable, Klaviyo (active-profile billing since Feb 2025), Braze (export gated; use Currents warehouse sync), Mailchimp, SendGrid (Email Activity API capped at 6 req/min in 2026), Brevo, Marketo
• Twilio API — Usage-based; messages 1/s per long code default, 10/s on toll-free

Video / meetings

• Loom (Atlassian-owned now; direction uncertain), Otter.ai (Business+ enterprise-gated), Fireflies.ai (GraphQL, Business/Enterprise), Calendly (200 req/min, 2k/hr), Zoom Cloud Recording + Transcripts (Cloud Recording paid Zoom plan required; download URLs short-lived signed JWTs), Microsoft Stream (now via Graph Drive/Site endpoints — actually stable place to integrate), Vimeo Enterprise

C) Integration Platforms & MCP

Model Context Protocol (MCP)

• Spec/site: modelcontextprotocol.io
• Reference repo: github.com/modelcontextprotocol/servers
• Official registry: registry.modelcontextprotocol.io (GA Sep 8 2025)
• Community catalogs: pulsemcp.com/servers (~15k indexed), mcp.so, mcpservers.org
• Governance: Transferred to Agentic AI Foundation in 2025 (Anthropic donation; stewards include GitHub, Microsoft, PulseMCP)

Critical 2026 shift: Streamable HTTP transport replacing SSE — Atlassian deprecating SSE June 30 2026; 25+ first-party remote servers reachable by URL.

Reference / official Anthropic servers

• Filesystem — sandboxed read/write/list. Best learning baseline. 5/5
• Fetch — URL → markdown. 5/5
• Git — repo manipulation. 4/5
• Memory — knowledge-graph persistent memory. 4/4
• Sequential Thinking — debated usefulness. 3/3
• Everything — reference/test server. Scaffolding only

First-party vendor MCP servers (remote HTTP, production-grade)

Most relevant for Vivly to bundle/federate:

1. GitHub (github/github-mcp-server) — most-installed MCP. Bundle priority: highest
2. Slack (official remote 2026) — channels/messages/threads/search
3. Notion (official) — pages/databases
4. Linear (official) — issues/cycles/projects
5. Atlassian (Jira + Confluence) — official remote; SSE → Streamable HTTP migration June 30 2026
6. HubSpot (official) — CRM read + write
7. Salesforce — via Salesforce MCP connector and Composio variants
8. Stripe (official) — payments, subscriptions, invoices (read-mostly by default)
9. Sentry (official) — errors, traces; "agent fixes from editor" pattern
10. Vercel (official) — deployments, env vars, builds
11. Supabase (official) — project + DB introspection
12. Neon (official) — serverless Postgres provisioning + query
13. Cloudflare (official) — Workers, KV, R2, DNS
14. PostgreSQL / MySQL / SQLite (reference) — DB read+exec
15. Google Drive (reference) — file search + read
16. Brave Search / Tavily / Exa — web-search MCPs
17. Puppeteer / Playwright — browser automation
18. Figma (official) — design file introspection
19. Docker Hub / Kubernetes — infra-side MCPs
20. Ahrefs / Skyvia / K2view — data/SEO MCPs trending 2026

Recommended Vivly MCP starter bundle: GitHub + Slack + Linear + Notion + Atlassian + Google Drive + Postgres + Filesystem + Fetch

Integration meta-platforms

Unified-API platforms (CRITICAL for Vivly)

Strategic Notes for Vivly

1. Unified-API layer choice is load-bearing. Realistic options for internal-side data:

   • (a) Nango (OSS, code-first) — best unit economics, self-host. Recommended primary.
   • (b) Merge — deepest normalization, expensive at scale
   • (c) MCP-first — let users bring their own server; you federate
   • Best path = Nango + MCP combo — Nango handles OAuth + delta sync; MCP handles per-user / per-agent surfacing.

2. MCP is now the integration distribution channel. Shipping Vivly as an MCP server is at least as important as having client integrations. The 25+ remote first-party servers (Slack/Linear/Notion/Atlassian/GitHub/HubSpot/Salesforce/Stripe/Sentry/Vercel/Supabase) are the de facto "internal data" plumbing — Vivly should compose them rather than re-implement.

3. Rate-limit blast zones to plan around:

   • Slack non-Marketplace history endpoint (1 req/min in 2026)
   • Atlassian points-based limits (March 2026)
   • Salesforce per-org daily API allocation (you eat customer's budget)
   • Microsoft Graph throttling (cannot be raised)
   • OpenAlex $1/day default credit

4. Public-data backbone for Vivly:

   • Scholarly: OpenAlex + Crossref + Unpaywall + Semantic Scholar + arXiv + PubMed = 95% of scholarly coverage
   • Knowledge: Wikidata + Wikipedia REST + OSM (Overpass/Nominatim) + Wolfram = general knowledge / geo / computation
   • IP: EPO OPS + USPTO ODP

   All free, mostly CC0/permissive.

5. Deprecation watch:

   • USPTO ODP Beta (May 29 2026)
   • Atlassian SSE MCP (June 30 2026)
   • JSTOR Constellate (already)
   • Wikidata Query Service partition split
   • NCBI PMC E-Utils migration (Feb 2026)