Browse MCPs & Agents
Loading MCPs and agents…
Loading MCPs and agents…
Ingest your documents, then ask questions against them. Retrieval runs server-side, so only the passages that matter come back — not the whole corpus.
+ $0.01 per tool call
Secure payment via Stripe. Cancel anytime.
NiceBot already runs retrieval-augmented chat for websites. This listing exposes the part underneath: a private document store you can fill from your agent and question from your agent.
The shape is deliberately small. Ingest a document, list what you have, ask a question, delete what is stale. No organisations, no domains, no dashboard step - the workspace is created on your first call and you are never asked to name it.
ask is the tool worth having. An agent can already read a document you hand it; what it cannot do is find the relevant passage across hundreds of them without pulling everything into context. Retrieval runs here, and only the passages that matter travel back, with their similarity scores and the answer they support.
When nothing in your documents matches, the response says so. It does not fall back on general knowledge and present it as though it came from your corpus.
Ask what your own documentation says, instead of grepping it
Give an agent your product docs so its answers match your product
Search a contract or policy set too large to paste into a prompt
Keep a research corpus an agent can consult across sessions
Ground an agent in internal runbooks rather than general knowledge
Retrieval runs server-side - only matching passages enter your context
Every answer comes back with the passages it came from
Says plainly when nothing matches, rather than improvising an answer
answer: false returns raw passages when you would rather reason yourself
Chunks overlap, so an answer spanning a boundary is not cut in half
Your corpus is private to your subscription and survives cancellation
No. Subscribing here is the whole setup. There are no organisations or domains to create - your workspace is made on your first call and you never name it.
For one document, do that. This is for when you have hundreds and do not know which one holds the answer. Retrieval finds the passage and only that travels back, which is cheaper and more accurate than filling a context window.
The response says so and invents nothing. Answers are generated only from passages retrieved out of your own corpus, and those passages come back alongside so you can check the work.
Ingesting costs 3 units because it embeds every chunk, and asking costs 1 because it runs a generation. Listing and deleting are free.
They are kept. Cancelling withdraws access; it does not delete the corpus you built. Resubscribe and it is where you left it.
No reviews yet. Be the first to share your experience.
Setup notes and replies. Ratings stay in the reviews above.
Subscribe using the button above
Create an API key in Dashboard
Add the MCP to Cursor, Claude, or your client
https://www.mcpchannel.ai/api/mcpOne URL for every MCP you subscribe to. Authenticate with your API key in an X-MCP-Market-Key header.