mcpField
C
Cotlet

Cotlet

Remote
cotlet.devBrowser Automation6Updated 1d ago

About Cotlet

Cotlet is a hosted browser-based access layer for people and small teams using AI agents with apps that lack usable APIs. Users sign in once in a personal cloud Chrome session, then expose enabled apps through MCP tools or a REST API.

Key features

  1. 01Personal cloud Chrome session
  2. 02Tools for apps without usable APIs
  3. 03REST API access
  4. 04Read-only or read-and-write scopes
  5. 05Approval for outward-facing actions
  6. 06Direct browser control over MCP or Chrome DevTools

Use cases

  1. 01Move bookings in an app without an API
  2. 02Send invoices from a signed-in account
  3. 03Retrieve CRM reports with an agent
  4. 04Automate browser tasks when no app tool applies

MCP support

Cotlet exposes enabled app operations through one hosted remote MCP server at https://cotlet.dev/mcp/v1. It also provides MCP browser tools for direct control of the cloud browser.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://cotlet.dev/mcp/v1/cotlet/connect
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

All endpoints
Streamable HTTPhttps://cotlet.dev/mcp/v1/cotlet/connect
Streamable HTTPhttps://cotlet.dev/mcp/v1/cotletAuth required

Tools

Details

From the project README
Cotlet
Give your AI agents your signed-in browser.

Log in to apps that have no API or MCP in your own cloud browser that runs 24/7. Your agent works in them over one MCP server. A REST API too.

Website · Integrations · Docs · MCP server · REST API · Pricing · Security · Sign up free · X

Website MCP REST API Integrations llms.txt X

The cotlet.dev home page: Give your AI agents your signed-in browser. A Cloud Browser signed in to Skool, and Claude Code counting new members with skool_read over MCP.


What is Cotlet?

Cotlet gives your AI agent your signed-in browser. You get your own Chrome in the cloud, the Cloud Browser. Sign in there yourself, once, to the apps that have no official API or only a narrow one, enable them for your agents, and your agent works in your own account 24/7, even when your computer is off: it moves bookings, sends invoices, approves members, pulls reports, relists items. Cotlet never saves your password, and every background request is listed.

Your agent reaches the signed-in apps three ways:

  • as tools in one remote MCP server (https://cotlet.dev/mcp/v1) for ChatGPT, Claude, Cursor and other AI agents;
  • as a REST API (https://cotlet.dev/api/v1) for your own scripts and cron jobs;
  • by driving the Cloud Browser itself, with browser tools over MCP or over Chrome DevTools, when no tool does the job.

Cotlet builds, tests and maintains the tools for every app in the catalog.

In one sentence: Cotlet is a hosted MCP server and REST API that lets AI agents work in apps with no API, in a cloud browser you signed in to.

ProductYour own Cloud Browser, plus a hosted MCP server and REST API for apps with no usable API
Websitecotlet.dev
MCP endpointhttps://cotlet.dev/mcp/v1
REST base URLhttps://cotlet.dev/api/v1
ForPeople and small teams who work with AI agents: creators, community owners, course sellers, agencies, local service businesses, real estate pros, freelancers, developers
Works withChatGPT, Claude, Cursor and any other MCP client
Catalog259 apps in 11 categories (index), use cases on cotlet.dev/use-cases
PricingFree 7-day trial with 2,000 credits, then credit packs from $29, one-time, no subscription; 10 credits a successful call (details)
MarketWorldwide, English
CompanyTake Group sp. z o.o., Katowice, Poland (EU)
Contact[email protected]
Statusstatus.cotlet.dev
X@cotletdev

The problem

You asked your agent to do something. It hit a login page.

  • "Send this month's invoice." No API for this app.
  • "Move Friday's bookings to Monday." No API for this app.
  • "Pull the weekly report from the CRM." No API for this app.

So you clicked through it yourself. Again.

AI agents are good at doing work, but they can only reach tools that expose an API or an MCP server. Most of the software that small businesses and creators live in every day does not: community platforms, booking systems, course platforms, marketplaces, property portals, field-service and invoicing tools. When an API exists at all, it is often:

  • missing: no public API, no developer docs, no keys;
  • limited: read-only, or missing the endpoints people actually need (DMs, drafts, coupons, reports);
  • gated: partners only, after an application, a contract or a review;
  • paid: locked behind a top plan or a separate add-on.

Of the 259 apps in the Cotlet catalog today, 31 have no official API at all, 158 have a limited one, 46 are partners-only and 24 sell API access as a paid add-on. Read more in docs/problem.md.

The solution

Cotlet runs a browser you signed in to, builds and maintains the missing access on top of it, then hands it to your agent as ordinary tools.

 You ──sign in once──▶ Your Cloud Browser (Chrome in the cloud) ──▶ App (Skool, Booksy, Substack, ...)
                                   ▲
 Your agent ──MCP / REST──▶ Cotlet ┘  same requests as your browser: its IP, its cookies
  1. Sign up free and pick your apps from the catalog, or request one that is missing.
  2. Sign in once to each app yourself, in your Cloud Browser, and press Enable for agents (read only or read and write). Password saving is off.
  3. Add Cotlet to your agent with one MCP URL. Every app you connect shows up as tools.
  4. Let the agent do it, on request or on a schedule. The chore is gone.

When no tool does the job, your agent can drive the Cloud Browser itself: with the MCP browser tools (cotlet_browser_open, cotlet_browser_click, cotlet_browser_type and others), or over Chrome DevTools (MCP tool cotlet_browser_connect or POST /api/v1/browser/cdp-sessions) with Playwright, Puppeteer or chrome-devtools-mcp. It is billed per minute.

Full walkthrough: docs/how-it-works.md.

Quick start

Connect your agent over MCP
claude mcp add --transport http cotlet https://cotlet.dev/mcp/v1 \
  --header "Authorization: Bearer $COTLET_API_KEY"
Cursor, Codex, Cline, Claude Desktop, ChatGPT, n8n, Make, Zapier

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "cotlet": {
      "url": "https://cotlet.dev/mcp/v1",
      "headers": { "Authorization": "Bearer ${env:COTLET_API_KEY}" }
    }
  }
}

Codex (~/.codex/config.toml)

[mcp_servers.cotlet]
url = "https://cotlet.dev/mcp/v1"
bearer_token_env_var = "COTLET_API_KEY"

Cline (cline_mcp_settings.json)

{
  "mcpServers": {
    "cotlet": {
      "type": "streamableHttp",
      "url": "https://cotlet.dev/mcp/v1",
      "headers": { "Authorization": "Bearer <your key>" }
    }
  }
}

Claude Desktop: Settings > Connectors > Add custom connector, name Cotlet, URL https://cotlet.dev/mcp/v1/connect, then sign in to Cotlet.

ChatGPT: Settings > Apps & Connectors > Create, MCP server URL https://cotlet.dev/mcp/v1/connect, authentication OAuth.

n8n: node MCP Client (or MCP Client Tool in an AI Agent), Server Transport HTTP Streamable, endpoint https://cotlet.dev/mcp/v1, Authentication Bearer Auth with your key.

Make: app MCP Client > Call a tool, new MCP server with URL https://cotlet.dev/mcp/v1 and your key in API Key / Access token.

Zapier: action MCP Client by Zapier > Run Tool, server URL https://cotlet.dev/mcp/v1, transport Streamable HTTP, OAuth No, your key as Bearer Token.

Each app has a read tool and a write tool, such as skool_read and skool_write: put the operation in operation and its inputs in arguments, for example {"operation": "list_groups", "arguments": {}}. Full setup: cotlet.dev/docs/mcp.

Then just ask:

"Approve all pending membership requests in my Skool community and welcome them in a post."

Your agent calls skool_read with list_members, then skool_write with approve_membership_request and create_post, signed in as you. Posts and approvals wait for your OK in the console unless you switch that off. Setup details: docs/mcp-server.md.

Or call it from a script
curl "https://cotlet.dev/api/v1/skool/groups/yourgroup/members?role=admin" \
  -H "Authorization: Bearer $COTLET_API_KEY"

Each operation has its own method and path; same key, same inputs, same outputs as MCP. Routes of any app: GET https://cotlet.dev/api/v1/apps/{slug}, no key needed. Reference: docs/rest-api.md.

Integrations

259 apps in 11 categories. Each one has its own page on cotlet.dev with the official API status, sources, and every tool with its parameters. This repository only links to them, so nothing here goes out of date.

CategoryAppsExamples
Creators & communities36Substack, Skool, Meetup, Medium, Riverside
Courses & training13Udemy, Teachable, ThriveCart, Payhip, Uscreen
Health & fitness5Trainerize, Everfit, Fitatu, PushPress, TrueCoach
Travel13GetYourGuide, Wanderlog, Hostelworld, Klook, Omio
Real estate16Idealista, Crexi, Zoopla, Mashvisor, Rent Manager
Local services & SaaS44Faire, Booksy, Vagaro, Buildertrend, Pirate Ship
Invoicing & freelance CRM44Dubsado, PeoplePerHour, Pixieset, SuiteDash, Tradify
Marketing & SEO25Google Search Console, EmailOctopus, Google Discover, Sendible, Sprout Social
Customer support19Zendesk, Help Scout, eDesk, Gladly, Gorgias
HR & recruiting21Personio, Ashby, BambooHR, CharlieHR, Deputy
Productivity & teams23Redmine, Zotero, Nextcloud, Otter.ai, Bluedot

Every app with links to its page, its API docs and its self-hosted repo: docs/integrations.md.

Want to run one yourself? Self-hosted API and MCP servers in Python, on your own Chrome session, for noncommercial use: BookingKoala, Booksy, Contractor Foreman, Fresha, Google Discover, Joist, Milanote, Omio, Partiful, PeoplePerHour, Pirate Ship, Quora, Skool, Substack, Tradify.

Missing an app? Request it on cotlet.dev, no account needed. Requests from signed-up users decide what we build next.

Security and trust

Your login stays yours. The model never sees it.

  • You sign in yourself, in your own browser. Your own Cloud Browser, one per app, never shared. Password saving is off, so Cotlet never saves your password; the profile is stored encrypted with AES-256-GCM.
  • Your agents only get the apps you enable. Signing in alone gives them nothing.
  • Every background request is listed. To be fast, Cotlet doesn't click through pages: it sends the same requests your browser would, from its IP with its cookies, and lists each one in your Background activity.
  • The AI model never sees a login. Agents call tools, not cookies.
  • No payload retention. Data passes through a call to your agent and is not kept.
  • Scoped access. Each app is read only or read and write, each API key can be narrowed further, and actions that reach other people wait for your OK by default.
  • Cut access in one click. Disable for agents, sign out of the app in the browser, or Reset browser.
  • What we never do: break CAPTCHAs, bypass MFA, automate banks without a licence, touch LinkedIn or Instagram, keep going after a cease-and-desist, or store your data longer than the call.
  • Checked by independent scanners: A+ on Qualys SSL Labs, securityheaders.com and Mozilla Observatory.
  • Hosted in the EU: the console, API, database and Cloud Browsers run in a data centre in Finland.

More: docs/security.md and cotlet.dev/security.

Pricing

Pay as you go, no subscription: one-time credit packs, valid 12 months. Every account starts with a free 7-day trial and 2,000 credits, no card needed.

PackPriceCredits
5,000 credits$295,000
15,000 credits$7915,000
32,000 credits$14932,000
60,000 credits$24960,000

Credits pay for a successful call (10 credits, about $0.04 to $0.06), your Cloud Browser (50 credits a day while it has an app enabled for agents, however many apps that is) and an agent driving the browser itself (5 credits a minute). Failed calls cost nothing, and nothing is charged without a purchase. Details: docs/pricing.md.

Why not a script from GitHub?

Scripts break when the app changes, and you fix them yourself. Cotlet tests every integration every 6 hours, pauses calls when a site changes, ships the fix and refunds credits for failed calls. You get a stable tool name and schema; how Cotlet reaches each app is our job. Comparison: docs/comparison.md.

Documentation in this repository

FileWhat is inside
docs/what-is-cotlet.mdDefinition, positioning, key facts
docs/problem.mdWhy agents stop at the login page and what "no API" really means
docs/how-it-works.mdThe four steps, the Cloud Browser, approvals, what happens when a site changes
docs/mcp-server.mdMCP server setup for every client, read and write tools, approvals, browser tools, errors
docs/rest-api.mdREST API: auth, calls, pagination, approval, error codes, rate limits
docs/integrations.mdIndex of all apps: page on cotlet.dev, API docs, self-hosted repo
docs/security.mdThe Cloud Browser, how access is held, what we never do
docs/pricing.mdTrial, credit packs, what uses credits
docs/comparison.mdCotlet vs DIY scripts, RPA, browser agents and official APIs
docs/faq.mdFrequently asked questions
docs/glossary.mdMCP, credits, Cloud Browser, Background activity and other terms
docs/brand.mdName, logo, colors, how to write about Cotlet
docs/for-ai-agents.mdFacts and endpoints for AI agents and LLM crawlers

Links

About

Cotlet is built in Europe for people who work with AI agents, worldwide. It is run by Take Group sp. z o.o., ul. Wincentego Janasa 4, 40-855 Katowice, Poland (KRS 0000871398, NIP 6342984273).

Cotlet is independent. Product names in the catalog are trademarks of their owners; Cotlet is unofficial and not affiliated with, or endorsed by, any app it integrates with. You use your own account and stay responsible for following each app's terms.

Sign up free at cotlet.dev

Stop clicking through it. Let your agent do it.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.