When GitBook’s pricing stopped making sense
Monday morning, 9:30 AM. A message pops up in your Slack. The product manager sends a screenshot of the GitBook billing page. The new price stares back: $65/month per site on the Premium plan, plus $12/month per person. Your five-person team runs two documentation sites. You do the math. What used to cost tens of dollars has jumped to over $300 monthly. The CTO replies with one line: “Let’s meet and find something else.”
This scene played out across dozens of teams in 2026. GitBook restructured its pricing in October 2024, moving from simple per-seat billing to a dual “site fee + seat fee” model. Premium runs $65/month per site plus $12/month per person. The Ultimate tier, which includes AI Assistant, hits $249/month per site. Solo developers running a single docs site barely notice. Teams maintaining separate developer documentation and help centers, with five people editing, suddenly face $300+ monthly bills. And GitBook’s focus has clearly shifted toward AI over the past two years: GitBook Agent, AI Assistant, Channels. Long-time users feel the product drifting away from their core need of quietly writing documentation.
The free tier limits you to one user and publishing only to gitbook.io subdomains. Want a custom domain? That starts at $65/month. Need team collaboration? Add $12 per person. The “free entry, paid features” model becomes an expanding fixed cost as teams grow.
Someone left a G2 review: “I just want to write docs, not buy an AI chatbot.”
So if not GitBook, what?
Mintlify: When documentation becomes an AI agent entry point
A startup in Hangzhou builds API middleware. Eight people. The product iterates fast. Their technical documentation lived on GitBook initially, worked fine, until the product manager asked: “Can our docs be read directly by Cursor? When users ask questions in their AI editor, can they pull answers straight from our documentation?”
That’s exactly where Mintlify comes in.
Han Wang and Hahnbee Lee founded Mintlify in 2022, graduated from Y Combinator S22, and sprinted forward. They raised $18.5M in a Series A led by a16z in 2024, then closed a $45M Series B in April 2026, hitting a $500M valuation. Total funding: $67M. They serve over 20,000 companies, and the client list reads like an AI unicorn roster: Anthropic, Cursor, Perplexity, Replit, ElevenLabs.
Mintlify’s core idea treats documentation as “knowledge infrastructure.” They don’t just help you write and publish docs. They care about this: when an AI agent needs to understand your product, can your documentation be accurately parsed and cited?
In June 2026, Mintlify made a bold move. They killed the middle-tier plans ($150-$550/month) and launched a clean two-tier pricing structure. The Starter plan is completely free: web editor, Git sync, custom domains, API Playground, AI assistant, writing agent, automation workflows, even an MCP Server. AI features bill by credit, with 5,000 free credits monthly and $0.01 per additional credit. Enterprise is custom pricing, adding SSO, permissions, and SLA.
The strategy is clear: let every team start at zero cost. Once they get hooked on AI features and burn through credits, revenue follows naturally. For most small to mid-size teams, 5,000 credits per month covers daily use. But if you lean hard on the AI writing agent (roughly 115 credits per run) or trigger the AI assistant frequently to answer user questions (about 23 credits each time), you might spend $100-$300 monthly.
What wins developers over is Mintlify’s DX. Documentation content lives as MDX in your Git repo. Change a line of code, submit a PR, docs update and deploy automatically. This fits perfectly into existing developer workflows, zero learning curve. It also has a built-in MCP Server, meaning your docs can be directly referenced by any AI tool supporting Model Context Protocol. As AI agents increasingly replace humans reading documentation, this capability has real strategic value.
Back to the Hangzhou team. They spent one afternoon migrating from GitBook to Mintlify. The next day they saw results in Cursor: users asking questions in the AI editor got answers citing their documentation’s code samples directly. This “documentation read by AI” experience is what Mintlify actually sells.
ReadMe: Turning API documentation into product experience
Another scenario. You lead developer relations at a Fintech company. Your API connects to over 300 clients. The most common question in your tech support channel every day: “How do I call this endpoint? What parameters do I fill? What does the response mean?” You don’t just need a documentation site. You need an “API playground” where customers can click twice and get it running.
ReadMe has been doing this since 2014, and by now it’s one of the most mature commercial solutions in API documentation. Intercom, Yelp, and other large clients choose it for a consistent reason: the interactive API documentation provides a self-service user experience that directly reduces technical support workload.
ReadMe’s killer feature is the “Try It!” button. Users can send API requests, view responses, and debug parameters directly on the docs page, no need to open Postman separately. Combined with request logs and usage analytics, you see which endpoints get called most, where errors spike, then optimize documentation accordingly. This closed loop from documentation to debugging to feedback is what separates ReadMe from other tools.
In 2026, ReadMe offers three tiers: Starter free (single project, basic API Reference, includes AI Dropdown and MCP Server), Pro $250/month (team collaboration, custom MDX components, AI Doc Linting), Enterprise $3000/month (multiple projects, SSO, audit logs, advanced AI features). The price isn’t cheap, but if your use case is an API product facing external developers, the analytics and interaction capabilities ReadMe provides are hard to replace.
ReadMe also added LLMs.txt and MCP Server support in 2026, letting API documentation be parsed and invoked not just by humans but by AI agents. This aligns with industry trends: more and more API integrations aren’t humans manually writing code, they’re AI coding assistants auto-generating it. If your API docs aren’t AI-friendly, you lose half your potential integrators.
ReadMe’s limitation is obvious. It’s designed for “external-facing API documentation,” not internal knowledge bases or general help centers. If you just want to write a few technical guides for your team, ReadMe is overkill, and the $250/month Pro entry price doesn’t make sense.
Docusaurus: Ultimate freedom for engineers
There’s another kind of team. Their documentation lead is an engineer. Fluent in React, comfortable with Markdown, sets up GitHub Actions without blinking. This team doesn’t need fancy SaaS dashboards. They want total control and zero cost.
Docusaurus was built for this team.
Meta open-sourced Docusaurus in 2017, originally to solve messy documentation across their open-source projects. By 2026, it has over 64,000 GitHub stars, supports 25+ languages for internationalization, and is used by projects like React Native, Jest, and Redux.
v3, released in 2023, fully embraced React 18 and MDX 3. The September 2025 v3.9 update introduced Algolia DocSearch v4 integration, supporting AI conversational search (AskAI), turning the search bar into an intelligent assistant that understands context and delivers complete answers. This means Docusaurus is no longer just a “static site generator.” It’s keeping pace with AI too, but using the open ecosystem approach. Algolia’s free Build tier offers 10,000 search requests monthly, plenty for open-source projects and small teams.
Docusaurus’s philosophy is “docs as code.” Articles go in the docs/ folder, configuration lives in docusaurus.config.js, compile to pure static HTML/JS, throw it on GitHub Pages, Vercel, or Netlify, and hosting costs nothing.
What’s the tradeoff? Barrier to entry. You need to know React, configure a Node.js 20+ environment, and integrate search, comments, and analytics yourself. For teams without frontend engineers, Docusaurus initial setup might take a day or two, and ongoing maintenance requires continuous engineering time. There’s no WYSIWYG editor. Product managers and technical writers can’t jump in and write content directly. Everything must go through Git commits.
Flip the angle, though, and that’s actually the strength. Documentation participates in version control and code review as part of the codebase, which actually improves quality. For engineering organizations where the culture is already “everything is a PR,” Docusaurus doesn’t introduce any new collaboration patterns. It just puts documentation under the same roof as code.
Archbee: For teams that need “good enough”
Not every team needs AI agent-level documentation platforms, and not every team has engineers willing to wrestle with static site generators. Some teams have simple needs: write docs, publish docs, team can edit together, customers can search for answers.
A mid-sized B2B SaaS company has a product team of a dozen people. Documentation is scattered across Notion, Google Docs, and an old Confluence space. They tried GitBook, liked it, but as the team grew and multiple product lines expanded, GitBook’s pricing started making finance frown. They ultimately chose Archbee.
Archbee is a small but solid documentation platform company. Total funding only $300-400K, but riding on product quality and word-of-mouth, they earned a 4.6/5 rating on G2 (118 verified reviews) and 4.7/5 on Capterra. Users repeatedly mention: fast, smooth editor, solid Markdown support, responsive customer service. It positions itself as an “all-stack technical documentation platform,” covering three scenarios: API docs, internal knowledge base, user help center.
Pricing also takes a pragmatic line: Growing plan $50/month (annual billing $40/month), Scaling plan $200/month (annual billing $160/month), Enterprise custom. No free tier, but there’s a 50% discount program for early-stage startups lasting two years, effectively $25/month to start. This is almost the lowest commercial documentation platform entry price on the market. Compared to GitBook’s “site fee + seat fee” double extraction, Archbee is more friendly to mid-sized teams.
Users repeatedly praise Archbee’s editor experience. Its AI assistant can answer user questions directly in documentation, reducing technical support pressure. Multi-product line docs can be organized through Space Links and Version Links, no need to spin up extra sites. Built-in search analytics let you see what users searched for, what they found, and which searches returned no results, providing data for documentation optimization.
Its shortcoming is brand awareness and ecosystem. Compared to GitBook and Mintlify’s high visibility in developer communities, Archbee is more of a “tool you know if you know” kind of thing. Also, no free tier makes it less friendly to individual developers and open-source projects. But for mid-sized commercial teams needing a stable, sufficient, no-fuss documentation platform, it might be exactly that “less is more” choice.
Core feature comparison
| Dimension | Mintlify | ReadMe | Docusaurus | Archbee | GitBook |
|---|---|---|---|---|---|
| Entry pricing | Free (AI pay-per-credit) | Free basic / Pro $250/mo | Fully free open source | $50/mo start | Free / Premium $65/site/mo + $12/person/mo |
| AI capability | Native AI assistant + writing agent + automation workflows | AI search + Doc Linting + MCP | Via Algolia AskAI integration | AI search assistant | AI Lens + AI Assistant (Ultimate+) |
| Git sync | Native MDX + GitHub | Supported | Native (code as docs) | Supported | GitHub/GitLab bidirectional sync |
| API Playground | Yes | Core feature, industry-leading | Needs plugin | Yes | Yes |
| Custom domain | Included free | Requires paid plan | Self-hosted, unlimited | Included | Premium+ ($65/mo) |
| Internationalization | Supported | Limited | Native support for 25+ languages | AI-assisted translation | Paid add-on |
| Deployment | Hosted SaaS | Hosted SaaS | Self-hosted (static site) | Hosted SaaS | Hosted SaaS |
| Fits team size | Startup to mid-size | Mid-size to enterprise | Any (needs engineering) | Small to mid-size | Startup to enterprise |
How to choose: Back to your actual scenario
Picking a documentation tool has no universal answer, only the answer that fits your current state.
If your team is building an API product with the goal of getting external developers up to speed quickly, ReadMe’s interactive experience and usage analytics are the most mature choice. The $250/month Pro plan isn’t expensive for a company with commercial API revenue. The time saved on technical support far exceeds that cost.
If your team has React engineers, values total control, and is budget-sensitive, Docusaurus is the only truly zero-cost solution. You don’t pay any SaaS monthly fees. The tradeoff is upfront setup time and ongoing maintenance labor. Open-source projects pick it almost by default.
If you’re running a growing SaaS product where docs need to reach users and be read by AI agents, Mintlify is the momentum pick for 2026. The free Starter plan lets you launch at zero cost. AI writing agent and automation workflows significantly lower documentation maintenance labor. Its client list is a signal: when Anthropic and Cursor both hand their docs to Mintlify, its AI ecosystem fit speaks for itself.
If you need a stable, sufficient, no-frills documentation platform that handles both internal knowledge base and external docs, Archbee’s value is high. $50/month entry price includes team collaboration, custom domain, AI search. It might not make Hacker News headlines, but it can quietly help you get documentation done well.
As for GitBook, it hasn’t gotten worse. The editor is still elegant, Git Sync still works well, G2 rating still sits at 4.8/5. But its pricing strategy and product direction are shifting: from a pure documentation tool, it’s becoming an AI knowledge management platform. If that’s exactly the direction you want, it’s still a good choice. But for teams that just want to quietly write documentation without paying extra for AI features, the market now has better-fitting alternatives.
The scariest part of choosing isn’t picking wrong. It’s staying stuck in a tool that no longer fits while hesitating. Documentation tool migration costs aren’t as high as you’d think. Most platforms support Markdown import. Moving content usually takes a day or two. The real difficulty is the moment you decide.
Why not spend half a day testing two candidates? Register a Mintlify Starter account, fork a Docusaurus template, or open an Archbee trial space. Move the three to five pages you maintain most often into each one. Let the team actually work through it. Whoever feels comfortable, whose workflow runs smoothest, the answer surfaces naturally.
Stay updated with our latest AI insights
