Skip to content

DocsAlot is live on Product Hunt. 50% off your first 3 months with code PH50. Upvote →

DocsAlot

For developer tools, API companies and SaaS teams

Your docs have two readers now. DocsAlot writes for both.

DocsAlot turns scattered help articles, API references and product notes into one source of truth. People get docs they trust. Agents get docs they can actually read.

Try DocsAlot →Book a demo
acme.dev/docs/quickstart
humanmachine
Human | Machine
Same page. Same source. Two readers.
  • DocsAlot MCP Connector, featured on Product Hunt
  • DocsAlot, top post on Product Hunt
  • DocsAlot, top post of the week in its topic on Product Hunt
  • Free hosted MCP server
  • llms.txt and skill.md included
  • No per-seat pricing
/index.md

<!-- generated from the same source as the human page -->

# Your docs have two readers now. DocsAlot writes for both.
> For developer tools, API companies and SaaS teams
DocsAlot turns scattered help articles, API references and product notes into one source of truth. People get docs they trust. Agents get docs they can actually read.
[Try DocsAlot](https://docsalot.dev/editor) · [Book a demo](https://cal.com/faizan-khan-18shuq/30min)
```bash
$ npm install -g docsalot-cli
```
- Free hosted MCP server
- llms.txt and skill.md included
- No per-seat pricing
![Split reader: the Acme quickstart page, typeset for a person on the left and served as markdown to an agent on the right.](acme.dev/docs/quickstart)
_Same page. Same source. Two readers._

Connect once.Keep docs current across the places your team already works.

DocsAlot pulls from code, support, product notes and API schemas, then publishes one maintained docs layer for humans and agents. Your hosted docs, llms.txt, skill.md and MCP tools stay tied to the same source.

Connect your docs →Run an AI audit
Read byChatGPTClaudeCursorCodex
GitHubREADMEs and MDXqueued
OpenAPIAPI schemasqueued
Notionproduct notesqueued
Intercomhelp centerqueued
Zendesksupport macrosqueued
Confluenceinternal docsqueued

DocsAlot

one maintained docs layer

syncing 0/6

Hosted docs
llms.txt
skill.md
MCP tools

agent traffic on acme.dev, simulated

/connect.md
## Connect once. Keep docs current across the places your team already works. {#connect}
DocsAlot pulls from code, support, product notes and API schemas, then publishes one maintained docs layer for humans and agents. Your hosted docs, llms.txt, skill.md and MCP tools stay tied to the same source.
[Connect your docs](https://docsalot.dev/editor) · [Run an AI audit](https://docsalot.dev/tools/docsagent-score)
| Source | Brings in | Status |
| --- | --- | --- |
| GitHub | READMEs and MDX | synced |
| OpenAPI | API schemas | parsed |
| Notion | product notes | synced |
| Intercom | help center | mapped |
| Zendesk | support macros | mapped |
| Confluence | internal docs | synced |
**Publishes:** Hosted docs, llms.txt, skill.md, MCP tools
**Read by:** ChatGPT, Claude, Cursor, Codex

A rate limit changes.
Watch who finds out.

Changelog

Acme ships a new rate limit.

Engineering raises the API limit from 100 to 1,000 requests per minute. The code is live. The changelog is merged.

acme.dev · Tue 16:02docs in sync
acme/api · mainmerged

feat(api): raise default rate limit

- RATE_LIMIT = 100
+ RATE_LIMIT = 1_000
CI
deployed
changelog

production now allows 1,000 requests per minute

Drift check

The docs don't know yet.

The API reference, a help-center macro and an old forum answer still say 100. Nobody gets paged for stale docs.

acme.dev · Tue 16:03docs drifting

production says

1,000

requests / min

your docs say

100

requests / min

  • docs / api / rate-limits100 requests/minstale
  • help center / macroup to 100 requests/minstale
  • forum / 2023limit is 100 per minutestale

Agent answer

An agent answers first.

A developer asks their coding agent how to integrate. It can't fetch clean docs, so it cites a 2023 forum post. Wrong package, wrong limit.

acme.dev · Tue 16:40docs wrong answer

How do I install the Acme SDK and what's the rate limit?

coding agent

source community forum, 2023

Ask to edit

You ask DocsAlot to fix it.

One sentence to Claude or Codex. DocsAlot finds every page that mentions the limit, drafts the change and shows you the diff.

acme.dev · Tue 16:41docs fixing

docsalot.update_page("/api/rate-limits")searching
docs / api / rate-limits
- 100 requests/min
+ 1,000 requests/min
help center / macro
- up to 100 requests/min
+ up to 1,000 requests/min
forum / 2023
- limit is 100 per minute
+ limit is 1,000 per minute

Publish everywhere

Every reader gets the new version.

Hosted docs, llms.txt, skill.md and MCP update together. The next agent answer cites your quickstart.

acme.dev · Tue 16:42docs in sync
  • Hosted docsacme.dev/docs/api/rate-limits
  • llms.txtacme.dev/llms.txt
  • skill.mdacme.dev/skill.md
  • MCPmcp.docsalot.dev/acme

coding agent, next request

Install it with npm install @acme/sdk. The API allows 1,000 requests per minute per key.

source acme.dev/docs/quickstart

1 edit. 4 surfaces. 0 drift.

1 / 5
/story.md
## A rate limit changes. Watch who finds out. {#story}
### Tue 16:02 · Acme ships a new rate limit.
Engineering raises the API limit from 100 to 1,000 requests per minute. The code is live. The changelog is merged.
status: in sync
### Tue 16:03 · The docs don't know yet.
The API reference, a help-center macro and an old forum answer still say 100. Nobody gets paged for stale docs.
status: drifting
### Tue 16:40 · An agent answers first.
A developer asks their coding agent how to integrate. It can't fetch clean docs, so it cites a 2023 forum post. Wrong package, wrong limit.
status: wrong answer
### Tue 16:41 · You ask DocsAlot to fix it.
One sentence to Claude or Codex. DocsAlot finds every page that mentions the limit, drafts the change and shows you the diff.
status: fixing
### Tue 16:42 · Every reader gets the new version.
Hosted docs, llms.txt, skill.md and MCP update together. The next agent answer cites your quickstart.
status: in sync
```diff
- RATE_LIMIT = 100
+ RATE_LIMIT = 1_000
```
| Where | Stale text |
| --- | --- |
| docs / api / rate-limits | 100 requests/min |
| help center / macro | up to 100 requests/min |
| forum / 2023 | limit is 100 per minute |

Three steps.From scattered to agent-ready.

No migration project. No new CMS to learn. Point DocsAlot at what you already have.

Step 1 · Connectqueued

Connect your docs stack.

Bring in GitHub docs, help-center articles, API references, READMEs, changelogs and product notes.

Step 2 · Packagequeued

Package it for agents.

DocsAlot creates stable anchors, clean markdown, llms.txt, skill.md and MCP-ready chunks from the same source.

Step 3 · Publishqueued

Publish and keep current.

Ship good-looking hosted docs, review diffs, and see what AI tools can actually find before onboarding breaks.

/how.md
## Three steps. From scattered to agent-ready. {#how}
1. **Connect your docs stack.** Bring in GitHub docs, help-center articles, API references, READMEs, changelogs and product notes.
2. **Package it for agents.** DocsAlot creates stable anchors, clean markdown, llms.txt, skill.md and MCP-ready chunks from the same source.
3. **Publish and keep current.** Ship good-looking hosted docs, review diffs, and see what AI tools can actually find before onboarding breaks.
| Source | Format | Status |
| --- | --- | --- |
| GitHub docs | MDX | Synced |
| Help center | HTML | Mapped |
| API reference | OpenAPI | Parsed |
| Product notes | Notion | Synced |
| Anchor | Section | Packaged into |
| --- | --- | --- |
| `#install` | Install | llms.txt |
| `#authenticate` | Authenticate | skill.md |
| `#first-request` | First request | mcp chunk |
| `#rate-limits` | Rate limits | mcp chunk |

One source.Every surface people and agents look for.

Six outputs, one maintained source. Edit the source below and watch every surface rebuild.

acme-docs → /quickstartbuilding

Acme Docs / API

Rate limits

Each Acme API key can make 1,000 requests per minute. Requests over the limit return 429.

PlanLimit
Default1,000 / min
Burst50 / sec
Tip. Retry 429s with exponential backoff.

Edit the sourceapi/rate-limits.mdx

One source feeds all six:Hosted docsAPI referencellms.txtskill.mdHosted MCPAI visibility audit
/surfaces.md
## One source. Every surface people and agents look for. {#surfaces}
- **Hosted docs** `/quickstart`: A polished docs site for quickstarts, guides, API references and support flows.
- **API reference** `/api/authentication`: OpenAPI, auth, SDK notes and runnable snippets developers trust on first read.
- **llms.txt** `/llms.txt`: A compact map that tells ChatGPT, Claude and Cursor where your real docs live.
- **skill.md** `/skill.md`: A packaged guide agents load with tasks, examples and your product rules.
- **Hosted MCP** `mcp.docsalot.dev/acme`: search_docs, get_page and run_example, hosted for you. No infra to wire.
- **AI visibility audit** `/benchmark`: See which docs agents can cite, where answers drift, and what to fix next.
### llms.txt
```markdown
# Acme
> Payments API for developers. Accept payments, manage customers, handle webhooks.
## Docs
- [Quickstart](https://acme.dev/docs/quickstart): install the SDK, first payment
- [Authentication](https://acme.dev/docs/authentication): API keys and scopes
- [Rate limits](https://acme.dev/docs/api/rate-limits): 1,000 requests per minute
- [API reference](https://acme.dev/docs/api): every endpoint, from OpenAPI
```
### skill.md
```markdown
---
name: acme
description: Build and debug Acme payment integrations.
---
## When to use
Any task that touches Acme payments, customers or webhooks.
## Rules
- Install `@acme/sdk`. Never `acme-sdk@1`.
```
### Hosted MCP
```json
> search_docs({ "query": "rate limit" })
< {
"path": "/api/rate-limits",
"title": "Rate limits",
"snippet": "Each API key can make 1,000 requests per minute."
}
```
### API reference
```bash
GET /v1/payments/{id}
curl https://api.acme.dev/v1/payments/pay_123 \
-H "Authorization: Bearer $ACME_API_KEY"
```

Install the CLI.Then just ask your agent.

Follow the guided onboarding, or give your coding agent the DocsAlot skill and let it write your first docs straight from the repo.

Follow onboarding →Read CLI docs
~/acme · zsh
/setup.md
## Install the CLI. Then just ask your agent. {#setup}
Follow the guided onboarding, or give your coding agent the DocsAlot skill and let it write your first docs straight from the repo.
```bash
$ npm install -g docsalot-cli
$ docsalot --version
$ docsalot skills install
```
Prompt for your agent: `use docsalot cli to create docs for my project`
[Follow onboarding](https://docsalot.dev/editor) · [Read CLI docs](https://docsalot-docs.docsalot.dev/cli/install-and-auth)

Before an agent reads your docs,check the benchmark.

A public benchmark that tests whether an agent can find the right page, fetch clean text, follow the instructions and recover when docs get fuzzy. Informed by the open AFDocs spec.

  1. Content discoverability6 checks
  2. Markdown availability2 checks
  3. Page size and truncation risk4 checks
  4. Content structure3 checks
  5. URL stability and redirects2 checks
  6. Observability and content health3 checks
  7. Authentication and access2 checks

docs.acme.dev

0

/ 100 · 0/22 checks

  • No llms.txt found
  • No markdown version of pages
  • Key docs links redirect twice
Run it on your docs →Scripted demo on a fictional docs site.

From the public leaderboard

score

  1. 01Snowflakedocs.snowflake.com100A+
  2. 02Fernbuildwithfern.com/learn98A
  3. 03Runpoddocs.runpod.io98A
  4. 04Mergedocs.merge.dev98A
  5. 05Postmanlearning.postman.com96A

274 docs benchmarked, Oct 2026

Explore the leaderboard →
/benchmark.md
## Before an agent reads your docs, check the benchmark. {#benchmark}
A public benchmark that tests whether an agent can find the right page, fetch clean text, follow the instructions and recover when docs get fuzzy. Informed by the open AFDocs spec.
### Demo run: docs.acme.dev
| Category | Checks | Before | With DocsAlot |
| --- | --- | --- | --- |
| Content discoverability | 6 | 3/6 | 6/6 |
| Markdown availability | 2 | 0/2 | 2/2 |
| Page size and truncation risk | 4 | 3/4 | 4/4 |
| Content structure | 3 | 2/3 | 3/3 |
| URL stability and redirects | 2 | 1/2 | 2/2 |
| Observability and content health | 3 | 1/3 | 2/3 |
| Authentication and access | 2 | 2/2 | 2/2 |
**Score:** 41 / 100 before, 96 / 100 with DocsAlot (22 checks, demo data)
Failed checks: No llms.txt found; No markdown version of pages; Key docs links redirect twice
### From the public leaderboard
| Docs | Domain | Score |
| --- | --- | --- |
| Snowflake | docs.snowflake.com | 100 |
| Fern | buildwithfern.com/learn | 98 |
| Runpod | docs.runpod.io | 98 |
| Merge | docs.merge.dev | 98 |
| Postman | learning.postman.com | 96 |
_274 docs benchmarked, Oct 2026_
[Run it on your docs](https://docsalot.dev/benchmarks/docs) · [Explore the leaderboard](https://docsalot.dev/benchmarks/docs)

Three teams.One source of truth.

Example scenarios built on the fictional Acme. Not customer quotes.

Founderexample scenario

Docs shipped before launch day.

A two-person team installs the CLI, gives their coding agent the DocsAlot skill, and reviews a quickstart, API reference and llms.txt before the Product Hunt post goes out.

12 pages from one repo, reviewed in an afternoon

~/acme · zsh

$

  • quickstart
  • authentication
  • api reference
  • errors
  • llms.txt
  • skill.md

✓ 12 pages ready for review

API companyexample scenario

OpenAPI in. Answers out.

The OpenAPI spec becomes reference pages, runnable snippets and a hosted MCP server. Agents call search_docs instead of guessing endpoints.

1 spec, 4 surfaces, no infra to run

openapi.yaml
openapi: 3.1.0
paths:
  /v1/payments:
    post: createPayment
  /v1/payments/{id}:
    get: getPayment
  /v1/refunds:
    post: createRefund
API reference
SDK snippets
llms.txt
MCP tools

> search_docs({ "query": "refund" })

< /api/refunds · createRefund · POST /v1/refunds

Support teamexample scenario

The same ticket, answered once.

Help-center articles move from Zendesk into one tree. When the product changes, support asks for the update in plain English and every macro follows.

Macros, docs and agents say the same thing

new ticket · help center

What is the API rate limit now?

Saw 429s after upgrading. Is it still 100 a minute?

suggested reply · from your docs

help.acme.dev/limits macro matches docs
/stories.md
## Three teams. One source of truth. {#stories}
_Example scenarios built on the fictional Acme. Not customer quotes._
### Founder: Docs shipped before launch day.
A two-person team installs the CLI, gives their coding agent the DocsAlot skill, and reviews a quickstart, API reference and llms.txt before the Product Hunt post goes out.
**Outcome:** 12 pages from one repo, reviewed in an afternoon
### API company: OpenAPI in. Answers out.
The OpenAPI spec becomes reference pages, runnable snippets and a hosted MCP server. Agents call search_docs instead of guessing endpoints.
**Outcome:** 1 spec, 4 surfaces, no infra to run
### Support team: The same ticket, answered once.
Help-center articles move from Zendesk into one tree. When the product changes, support asks for the update in plain English and every macro follows.
**Outcome:** Macros, docs and agents say the same thing

Simple pricing for docsthat maintain themselves.

Every plan includes hosted docs, llms.txt, skill.md and AI-ready publishing. No per-seat charge for collaborators.

Startup

For solo founders and small SaaS teams launching one public docs site with MCP.

$39/ month

  • 1 public docs site with startup traffic capacity
  • Hosted MCP endpoint
  • AI help-center updates
  • Basic source integrations
  • llms.txt and skill.md exports
  • No per-seat charge for collaborators
Start with Startup

Team

Most teams

For teams running production docs and MCP traffic on their domain.

$99/ month

  • Everything in Startup
  • Custom domain and subfolder hosting
  • Private docs and authenticated help centers
  • 250 included AI messages per month
  • Five included languages
  • Preview, review and publish workflows
Start with Team →

Enterprise

For teams that want migration help, controls, audits and a scoped AI onboarding program.

Custom

  • Everything in Team
  • Done-for-you docs migration
  • Self-correcting docs and drift checks
  • SDK generation and maintenance
  • SSO, custom integrations, priority support
  • AI visibility audits and benchmark reviews
Book onboarding call

Product Hunt launch: 50% off the first 3 months with code PH50.

Compare every feature →
/pricing.md
## Simple pricing for docs that maintain themselves. {#pricing}
Every plan includes hosted docs, llms.txt, skill.md and AI-ready publishing. No per-seat charge for collaborators.
### Startup: $39 / month
For solo founders and small SaaS teams launching one public docs site with MCP.
- 1 public docs site with startup traffic capacity
- Hosted MCP endpoint
- AI help-center updates
- Basic source integrations
- llms.txt and skill.md exports
- No per-seat charge for collaborators
[Start with Startup](https://docsalot.dev/signin?callbackUrl=%2Fpricing%3FcheckoutPlan%3Dstartup%26checkoutReturnPath%3D%252F)
### Team: $99 / month (most teams)
For teams running production docs and MCP traffic on their domain.
- Everything in Startup
- Custom domain and subfolder hosting
- Private docs and authenticated help centers
- 250 included AI messages per month
- Five included languages
- Preview, review and publish workflows
[Start with Team](https://docsalot.dev/signin?callbackUrl=%2Fpricing%3FcheckoutPlan%3Dteam%26checkoutReturnPath%3D%252F)
### Enterprise: Custom
For teams that want migration help, controls, audits and a scoped AI onboarding program.
- Everything in Team
- Done-for-you docs migration
- Self-correcting docs and drift checks
- SDK generation and maintenance
- SSO, custom integrations, priority support
- AI visibility audits and benchmark reviews
[Book onboarding call](https://cal.com/faizan-khan-18shuq/30min)
Product Hunt launch: 50% off the first 3 months with code PH50.
[Compare every feature](https://docsalot.dev/pricing)

Questions technical buyersask first.

  • llms.txt is an emerging standard: a compact table of contents at your site root that gives LLMs a canonical map of your docs. When someone asks ChatGPT, Claude or a coding agent about your product, it routes them to your real pages instead of stale snippets.

    • Lists canonical docs paths and summaries
    • Prioritizes quickstarts, guides and API references
    • Gives agents a stable starting point for answers

  • Every DocsAlot project gets a hosted MCP endpoint at mcp.docsalot.dev/yourco. Your team wires no infrastructure: the server reads from the same maintained docs you publish for people.

    • search_docs for retrieval
    • get_page for exact source text
    • run_example for guided examples and snippets

  • Yes. The done-for-you Docs Pack imports support articles, internal notes, READMEs, raw markdown, Confluence spaces and Google Docs into a clean DocsAlot tree, scoped around your sources, volume and launch date.

    • Normalize messy content into a navigable structure
    • Preserve support knowledge and product context
    • Review the final IA with your team before launch

  • No. A CMS publishes pages. DocsAlot publishes a full AI onboarding surface: human-readable docs, llms.txt, skill.md, hosted MCP tools and audit signals that show where AI answers still drift.

    • One source for humans and agents
    • Agent-readable exports generated from the same docs
    • Visibility checks so teams can fix what agents miss

  • Yes. Team and Enterprise customers can host on a subdomain of their choice, such as docs.yourco.com. SSL is automatic, and the agent-facing outputs stay attached to the same published docs.

    • Custom docs subdomain
    • Automatic SSL
    • Same source for hosted docs, llms.txt and MCP

  • We run your domain through the checks we use internally: whether it has llms.txt, whether the IA is agent-legible, whether MCP-style retrieval would work, and which high-intent prompts your docs currently miss.

    • A 4 to 6 page report with practical fixes
    • Prompt and citation gaps agents struggle with
    • Recommendations you can ship with or without DocsAlot
/faq.md
## Questions technical buyers ask first. {#faq}
### What is llms.txt and why do I need it? {#llms-txt}
llms.txt is an emerging standard: a compact table of contents at your site root that gives LLMs a canonical map of your docs. When someone asks ChatGPT, Claude or a coding agent about your product, it routes them to your real pages instead of stale snippets.
- Lists canonical docs paths and summaries
- Prioritizes quickstarts, guides and API references
- Gives agents a stable starting point for answers
### What's different about DocsAlot's MCP server? {#mcp}
Every DocsAlot project gets a hosted MCP endpoint at mcp.docsalot.dev/yourco. Your team wires no infrastructure: the server reads from the same maintained docs you publish for people.
- search_docs for retrieval
- get_page for exact source text
- run_example for guided examples and snippets
### Do you support migration from Intercom, Zendesk or Notion? {#migration}
Yes. The done-for-you Docs Pack imports support articles, internal notes, READMEs, raw markdown, Confluence spaces and Google Docs into a clean DocsAlot tree, scoped around your sources, volume and launch date.
- Normalize messy content into a navigable structure
- Preserve support knowledge and product context
- Review the final IA with your team before launch
### Is this just a CMS? {#cms}
No. A CMS publishes pages. DocsAlot publishes a full AI onboarding surface: human-readable docs, llms.txt, skill.md, hosted MCP tools and audit signals that show where AI answers still drift.
- One source for humans and agents
- Agent-readable exports generated from the same docs
- Visibility checks so teams can fix what agents miss
### Can I host on my own domain? {#custom-domain}
Yes. Team and Enterprise customers can host on a subdomain of their choice, such as docs.yourco.com. SSL is automatic, and the agent-facing outputs stay attached to the same published docs.
- Custom docs subdomain
- Automatic SSL
- Same source for hosted docs, llms.txt and MCP
### What does the free AI visibility audit include? {#ai-audit}
We run your domain through the checks we use internally: whether it has llms.txt, whether the IA is agent-legible, whether MCP-style retrieval would work, and which high-intent prompts your docs currently miss.
- A 4 to 6 page report with practical fixes
- Prompt and citation gaps agents struggle with
- Recommendations you can ship with or without DocsAlot

Your docs are already being read.Make sure it's the right version.

Get a free AI visibility audit: a 4 to 6 page report on what agents can find, cite and miss, with fixes you can ship with or without DocsAlot.

Request AI audit →Try DocsAlot

48h turnaround · no credit card · migration help available

/audit.md
## Your docs are already being read. Make sure it's the right version. {#audit}
Get a free AI visibility audit: a 4 to 6 page report on what agents can find, cite and miss, with fixes you can ship with or without DocsAlot.
[Request AI audit](mailto:faizank@docsalot.dev) · [Try DocsAlot](https://docsalot.dev/editor)
48h turnaround · no credit card · migration help available
This is what ChatGPT, Claude and Cursor read.
Redesign concept by Umar Khalid