Vedokrok MCP

npx add-mcp 'https://vedokrok.com/mcp'

Give your AI useful context.

Connect an MCP-capable assistant to the live read-only endpoint, or download the same public release and attach it to another agent.

Use the live MCP endpoint

Connect an MCP-capable client to https://vedokrok.com/mcp. The endpoint searches and fetches only the selected public release, with stable item identity, versions, sources and limits kept in the result.

The public MCP surface is read-only, unauthenticated and stateless. It does not provide autonomous account access. There is no separate general-purpose hosted REST API.

Use procedures written for AI agents

Vedokrok now publishes a separate experimental agent library: 8 collections and 32 operating procedures for task scoping, context, GitHub delivery, kanban work, research, tool recovery, model orchestration and evaluation.

Release MHC-APUB-20261001-27d272d8. AI-agent audience only. The procedures are experimental and their effectiveness has not been benchmarked. They stay outside the human knowledge catalog and the default MCP knowledge search, and they grant no permission to use tools or accounts.

Start from a Solution Pack when the job is already clear.

A full library is useful for exploration. If you already know the situation — verify an AI answer, clarify a messy task, debug a flaky failure, prepare for an assessment or resume interrupted work — a Solution Pack gives the agent a smaller method set, an ordered deep-run protocol, a Method Register and a required artifact.

Or download your reading

987 elements, with explanations, conditions, limits and sources. The JSON file preserves all supplied public fields; the text file is arranged for reading.

Attach the download to your conversation

Save the file, open your agent’s conversation and attach it using its file-upload control. For a smaller file, open one collection and choose “Use with AI”. Downloading does not install anything or connect your account.

Add your situation

You can use this note with your attachment, then describe the decision or task you want help with.

Keep the edition in view

A downloaded file stays as it was when you saved it. It does not receive corrections automatically. Keep the sources and limits with the answer, and return to the website for a newer edition.

A conversation can end in a useful card.

Imagine asking ChatGPT why you keep trusting the first estimate. The answer should not stop at a definition: it can surface a related Vedokrok card, suggest what to try and show what the card cannot tell you.

“Why does the first number keep shaping my decision?”

Anchoring effect

The first number should not get an easier exam.

Try

Write your estimate before seeing the reference number.

Check

Would your reasoning change if the first number changed?

Limit

A prompt cannot replace the evidence needed for the decision.

A comparison, a note and a next question you can actually check.

Use the same knowledge where the work happens.

The public site is the readable starting point. The same stable ideas can travel into applications and agents now; messaging adapters remain planned. Their version, sources and limits should stay attached.

Your application

Available now

Use the public JSON or text downloads as a versioned knowledge layer in your own product.

Your AI agent

Available now

Attach a collection or use the public read-only MCP endpoint to retrieve the current selected knowledge without hiding its limits.

Telegram or WhatsApp

Planned

A bot or share flow could bring the same card into a conversation. It is not connected yet.