Every page as markdown
Send Accept: text/markdown to any karobar.io page and you get a markdown version of it instead of HTML: the page text, headings, lists, tables and links, without menus or product mock-ups. Browsers and search engines still get HTML. You can also open the markdown directly by adding /index.md to a page address.
curl -H "Accept: text/markdown" https://karobar.io/pricing
curl https://karobar.io/guides/take-orders-from-instagram-dms/index.mdEvery page also has a Link header pointing at its markdown version, the API catalog, this page, llms.txt and the sitemap.
Site data API
Read-only JSON about Karobar, rebuilt from the website on every deploy, so it always matches what the pages say. No key, no sign-up. The OpenAPI description is at /openapi.json and the catalog (RFC 9727) at /.well-known/api-catalog.
| Endpoint | What it returns |
|---|---|
| /api/v1/company.json | Company details, office, support and sales contacts |
| /api/v1/plans.json | Plans, monthly prices in USD, AI credits, users, contacts, accounts per channel, free trial |
| /api/v1/features.json | Every feature with its status and page links |
| /api/v1/platforms.json | The platforms Karobar connects to |
| /api/v1/guides.json | Guides; /api/v1/guides/{slug}.json adds the full text |
| /api/v1/comparisons.json | Comparisons; /api/v1/comparisons/{slug}.json adds the full text |
MCP server
A public, read-only Model Context Protocol server at https://karobar.io/mcp (Streamable HTTP, stateless, JSON responses, no authentication). Its card is at /.well-known/mcp/server-card.json.
get_plans: plans, prices and limitslist_featuresandget_feature: features with status, one feature page as markdownlist_platforms: the connected platformssearch_guidesandget_guide: find a guide, read it as markdownget_comparison: Karobar compared with respond.io, WATI or SleekFlowget_contact: company details and contacts
curl -s https://karobar.io/mcp -H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_plans","arguments":{}}}'A2A site guide
An Agent2Agent (A2A) endpoint at https://karobar.io/a2a that answers message/send requests. It is rule-based, not a language model: it recognises questions about plans, features, guides, comparisons and contacts, answers from the same site data, and links the page to read. Its card is at /.well-known/agent-card.json.
curl -s https://karobar.io/a2a -H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"message/send","params":{"message":{"role":"user","messageId":"1","parts":[{"kind":"text","text":"compare Karobar and WATI"}]}}}'Skills and discovery files
- /.well-known/agent-skills/index.json: agent skills for pricing questions, finding a guide, comparing tools, starting a trial and using the MCP server
- /.well-known/ai-catalog.json: one manifest listing the MCP server, the A2A guide, the API, the skills and llms.txt
- /llms.txt and /llms-full.txt: a summary of the site and the full text of every page
- /robots.txt: every crawler and agent is allowed, with content signals for search, AI answers and training
- WebMCP: on every page, browsers that support it get tools to read the plans, find a guide, open a page, start the free trial and get contacts
Accounts and authentication
Everything above is public and needs no credentials. Karobar accounts are created by people at app.karobar.io, and agents cannot register accounts or act inside a customer's Karobar account. Details are in auth.md.
What is not offered, and why
- OAuth discovery and protected-resource metadata: Karobar does not run an OAuth authorization server, and the public endpoints need no authorization, so there is nothing true to publish.
- Web Bot Auth: this is for sites that operate their own crawlers or agents that sign requests. Karobar does not.
- Agent commerce (ACP, UCP, x402 and similar): Karobar is bought by signing up and choosing a plan in the app, not through an agent checkout.
Limits and contact
The MCP and A2A endpoints are limited per IP address (bursts are allowed, then about 30 requests a minute) and return 429 when the limit is reached. Static files (markdown, JSON) are not limited. Please put your agent's name in the User-Agent header. Questions: support@karobar.io.