# Dc · Docs & Help — element 36 of 58 > Let the product explain itself. Turns questions into self-service answers. - **Group:** 8 · Support & Success - **Necessity:** Essential - **Price band:** $ · under $30/mo - **Maturity:** Stable - **Edition:** v2026.Q3 · verified 2026-09-13 ## Leading tools (v2026.Q3) - **Mintlify** — the agent-native default - **GitBook** — the classic, trust layer - **Fern** — spec → docs + sdks - **ReadMe** — api docs veteran - **Docusaurus** — open-source default ## Our take Your docs are now read by more agents than humans. Structure accordingly — it feeds your support element directly. ## Combines with Sa, Kw ## The top 5 — deep dossier (verified 2026-09-13) Mintlify, and 2026 is the year the reason changed: docs are now agent infrastructure. Across Mintlify-hosted sites, agents generated 66% of measured docs traffic in July 2026 (213M agent requests vs 105M human page loads), and Mintlify converted that shift into a $45M a16z Series B at $500M (Apr 2026) and $21.4M ARR. GitBook wins when non-technical teammates write half the docs and you want its trust/review layer; Fern when SDKs and docs should fall out of one API spec (now Postman-owned); ReadMe when an enterprise API program wants the 10-year veteran; Docusaurus when you want free, open-source, total control and are willing to bolt on the agent layer yourself. 1. **Mintlify** (Mintlify (YC W22)) — Starter free (5 seats, MCP server, API playground) · Pro paid/deployment + 10k AI credits, $0.01/credit overage · Enterprise custom. Best for: The default docs element for an AI-led startup — agent-readable by default (llms.txt, .md routes, hosted MCP), with the data to prove agents actually use it. Why: The category's center of gravity: 5,000+ hosted docs sites, Anthropic/Claude Code among its docs customers, $10.6M→$21.4M ARR in seven months (Jul 2026), and a $45M Series B at $500M led by a16z (Apr 14, 2026). Its July 2026 traffic report is the defining dataset of the era — agents at 66% of docs traffic and climbing — and Mintlify Index (Aug 6, 2026) turns those 5,000 sites into one retrieval layer for coding agents. Watch: Pro price is animated off the pricing page — budget clarity requires a sales touch; credit-metered AI features add bill variance. Its own benchmark data doubles as marketing; the 66% figure is Mintlify-network-only, not the whole web. [https://mintlify.com](https://mintlify.com) 2. **GitBook** (GitBook) — Free · Premium $65/site/mo · Ultimate $249/site/mo (+$12/user) · Enterprise custom (annual = 2 months free). Best for: Teams where PMs, support, and engineers all edit docs — visual editor plus Git sync, with llms.txt and an MCP server on every tier including free. Why: The classic, repositioned smartly for 2026: 2M+ users, and its pitch — "AI made docs easy to write. Not easy to trust." — attacks the failure mode agent-era docs actually have, drift and amplified errors. LLM optimization (llms.txt, llms-full.txt, .md routes) and an auto-generated MCP server ship on all tiers; the GitBook Agent detects stale content and AI Insights shows where humans and agents get stuck. Watch: The good AI (AI Assistant, Insights, adaptive content) is gated behind the $249/mo Ultimate tier, and per-site pricing stacks up across multiple products. No published traffic data comparable to Mintlify's; less gravitational pull among AI-native startups. [https://www.gitbook.com](https://www.gitbook.com) 3. **Fern** (Birch Solutions → Postman (acquired Jan 2026)) — Hobby free · Team $150/mo (annual) · Enterprise custom; AI credits tiered (250/1,000/custom). Best for: API-first companies that want docs, SDKs in 9 languages, llms.txt, and an MCP server all generated from one OpenAPI/AsyncAPI/gRPC spec. Why: The spec-to-everything play: one API definition yields reference docs, client SDKs, Ask Fern answers (890k+ served), and a queryable MCP server. Powers ElevenLabs (1.5M unique docs users/mo), Auth0, Webflow, Square. Postman acquired it on Jan 8, 2026 with brand and roadmap kept intact — plugging Fern into a 500k-company distribution machine. Watch: Acquisition risk cuts both ways: Postman promises independence, but roadmaps bend toward owners. Guide-style, non-API documentation is weaker than Mintlify/GitBook; customer-count claims vary wildly (homepage "8,000+ customers" vs Postman's "over 200 customers" — the latter is likely paying). [https://buildwithfern.com](https://buildwithfern.com) 4. **ReadMe** (ReadMe (YC W15)) — Starter free · Pro $250/mo · Enterprise $3,000+/mo · Ask AI add-on $150/mo. Best for: Enterprise API programs that want a decade-proven platform — interactive API explorer, git-native review flow, and agent read/write access to docs via MCP. Why: The API-docs veteran (NVIDIA, Amazon, PagerDuty) that re-platformed for the agent era: llms.txt and MCP servers on every tier including free, an AI Writer that proposes doc updates from your PRs, and MCP write access so Claude Code or Cursor can update docs directly from CI. "Documentation built for agents" is now the homepage headline. Watch: Priciest entry of the top five once you leave the free tier — $250/mo plus $150/mo for Ask AI — and enterprise starts at $3k/mo. Independent and profitable-quiet rather than momentum-loud; less pull with new AI-native startups than Mintlify. [https://readme.com](https://readme.com) 5. **Docusaurus** (Meta (open source, MIT)) — Free, MIT-licensed · you pay only for hosting (Vercel/Netlify/GitHub Pages). Best for: Engineer-written docs with zero platform spend and total control — React/MDX, versioning, i18n, and a community llms.txt plugin. Why: The open-source default: 65.1k GitHub stars, 12.2k dependent projects, actively shipped by Meta (v3.10.1, Apr 30, 2026). Everything the SaaS platforms charge for — versioning, localization, search — is free; the agent layer (llms.txt generation, MD routes) arrives via community plugins listed on llmstxt.org. Watch: You are the platform team: no hosted MCP server, no AI assistant, no agent analytics out of the box — each is a plugin or a build-it. Docs-site maintenance is exactly the kind of undifferentiated work an AI-led startup should be buying, not building. [https://docusaurus.io](https://docusaurus.io) ### How to choose - If You're an AI-native startup shipping your first docs site this quarter → Mintlify Starter — free, and llms.txt, .md routes, MCP server, and API playground are defaults, not add-ons. You inherit the network Claude Code's own docs team uses. - If Half your doc edits come from support, PMs, or founders who won't touch MDX → GitBook — the visual editor plus Git sync bridge, and its stale-content agent guards the accuracy problem AI writing creates. Budget for Ultimate if you want the AI features. - If You sell an API and are hand-maintaining SDKs alongside docs → Fern — one spec generates both, plus an MCP server agents can execute against. Price in the Postman-ownership question before a multi-year commitment. - If Docs traffic and agent 404s matter to you (they should) → Ship llms.txt and markdown routes whatever platform you choose — Mintlify's 2,400-run benchmark (Jul 2026) showed a single llms.txt link cuts agent 404s by ~90% (2.23 → 0.11 per run). - If Budget is zero and your engineers already write markdown in the repo → Docusaurus with the llms.txt plugin — then reevaluate the day docs maintenance shows up in sprint planning, because that's work the paid platforms have automated. ### The field (17 more) kapa.ai, Inkeep, Document360, Archbee, Redocly, Scalar, Bump.sh, Theneo, Read the Docs, MkDocs Material, VitePress, Stoplight (acquired), Fern (independent) (acquired), Trieve (acquired), Slate (dead), Zendesk Guide / Knowledge, Intercom Articles Full dossier data: https://elems.ai/e/dc.json --- Source: [elems.ai](https://elems.ai/e/dc.html) — the periodic table of the AI-led startup. Data: https://elems.ai/elements.json (CC BY 4.0, cite elems.ai).