Knowledge bases for AI agents
Collections your agent retrieves from. Add documents, ingest them into chunks and embeddings, and search by meaning before answering.
Copy the setup prompt.
Your agent does the rest.
Fetch and execute the appropriate instructions to set me up for superagnt_ from https://mcp.superagnt.com/agent-setup/prompt.md» once it is connected, run this so your agent can switch the source on itself.
agnt_tools_enable({ families: ['knowledge'] })Everything your agent can do with it
agnt_knowledge_create_collection makes the retrieval unit an agent binds to, with chunk size and overlap set per collection.
agnt_knowledge_add_document registers a source. agnt_knowledge_ingest chunks and embeds it.
Defaults of 512 tokens with 64 overlap suit prose. Raise the chunk size for reference tables and dense technical documents, where a small chunk splits a row from its header.
agnt_knowledge_search returns the best-matching chunks with their source document and a similarity score.
Search a fresh collection yourself to check it answers the questions the agent will face, before binding it to anything customer-facing.
List collections and the documents inside them, so you can see what an agent is actually retrieving from.
A collection can carry a refresh schedule, so a source that changes upstream gets re-ingested instead of going stale.
tool names06as your agent sees them
agnt_knowledge_create_collectionagnt_knowledge_add_documentagnt_knowledge_ingestagnt_knowledge_searchagnt_knowledge_list_collectionsagnt_knowledge_list_documentsThree jobs it can start on today
Answer from your docs, not from training data
Load product documentation into a collection. The support agent searches it before replying and cites the document each chunk came from.
Policy the agent has to follow
Put pricing rules, discount policy and contract terms in a collection, so an agent checks the current rule instead of repeating what it saw in an old thread.
Verify retrieval before you ship
Run the questions you expect through agnt_knowledge_search first. If the right chunks do not come back, fix the chunking or the source before an agent is answering customers.
What comes back
Your agent reads this response directly, with no parsing layer of your own in between.
https://mcp.superagnt.com/mcptool callagnt_knowledge_search{
"collection_id": "col_2f7a",
"query": "what is the refund window on annual plans",
"top_n": 2
}{
"chunks": [
{
"document_id": "doc_51b8",
"document_title": "Billing policy",
"similarity": 0.79,
"content": "Annual plans may be refunded in full within 30 days of the renewal date..."
},
{ "...": "1 more chunk" }
]
}What it costs to run
Platform tools carry no per-tool price. What you pay is the AI and credit usage the work causes, on the same workspace balance every other tool draws from.
Each run is itemized, so you can see which agent spent what and on which call.
how credits work →Before you wire it up
01How do I turn this on?
The knowledge family is elevated, so agnt_tools_enable refuses it. A person turns it on from the dashboard MCP page, because these tools change workspace configuration rather than only reading data.
02What chunk size should I use?
Start with the 512 token default and 64 overlap for prose. Raise it for reference tables and dense technical documents, where a small chunk separates a row from the header that gives it meaning.
03How do I know retrieval is working?
Call agnt_knowledge_search with the questions the agent will face and read the chunks and similarity scores that come back. That is far faster than debugging it through a conversation.
04Can a source stay current?
Yes. A collection can carry a refresh schedule, so a document that changes upstream is re-ingested instead of quietly going stale.
Point your client at https://mcp.superagnt.com/mcp and your agent has this, plus everything else on one balance.