The documents your agents can read
A Library holds your organization's documents so a granted agent can read them during a task without writing retrieval code.
A Library holds your documents so an agent can read them during a task. It covers document search, agent knowledge, knowledge base, and retrieval-augmented generation.
- An agent gains four tools: list, find by meaning, match exact string, and read lines.
- A resource grant names a subject and controls allow or deny access to a Library.
- A refusal grant beats every allow, and no agent widens its own reach.
- Built-in viewer and editor roles read every Library in the tenant.
Summaries were generated by AI. Generative AI is experimental.
A Library holds your organization’s documents. You upload a file once, grant a Library to an agent, and that agent can read the file during a task. You write no retrieval code: no text splitting, no embedding job, no vector database.
What an agent gains
Grant a Library and the agent gets four tools that reach every document in it:
- List what is there.
- Find the passage—the piece of a page—that means what the task asked, across your working languages.
- Match an exact string, such as an invoice number or an error code, literally.
- Read back the lines it needs to quote.
Those four tools let an agent inspect a Library instead of relying on ranked search alone.
What you keep
A resource grant controls which subject reaches a Library while a task runs. It names an agent, a person, or a group. Its effect is
allow or deny. This section calls a deny grant a refusal, and a refusal beats every allow. You decide once, in one
place, which agent reaches which Library, and no agent widens its own reach while a task runs.
A grant does not isolate Library content from people who hold tenant-wide roles. The built-in viewer and editor roles read every Library in the tenant, so use a separate tenant to keep document content away from a person. Access and isolation explains what those roles can read and when to use another tenant.
Other names for this capability
Other platforms sell this capability under other names. A reader who arrives looking for document search, agent knowledge, a knowledge base, a vector store, or retrieval-augmented generation (RAG) is in the right section. Two differences are worth stating early, because they change how you design a corpus:
- The exact-string match is its own tool, not a setting inside search. An invoice number is found literally, not approximated, because the agent asks for an exact match instead of asking for a relevant one.
- There is no query-time attribute filter. The Library is the scope. Put the documents that answer one kind of question in one Library, and grant that Library to the agent that asks it.
Choose a Library guide
- What agents can do with a Library—see how each tool matches, how to read an answer that stopped early, and how a document becomes searchable.
- Design your Libraries—choose how many Libraries to create, and which agent reaches each.
- Access and isolation—grant, refuse, and review who reaches a Library.
- Security and data handling—review encryption, malware scanning, deletion, and records.
- Limits—look up sizes, ceilings, windows, and what the platform meters.
- Best practices—prepare documents that index well, and keep answers complete.
- Patterns—the practical cases a team meets with real documents, and the method for each.
Manage libraries covers the client calls that create a Library and upload a document, and Attach documents covers the files one task carries.
When a document will not load, Errors and reliability lists each failure code and its recovery.
When a ready document gives a wrong or incomplete answer, start with
Troubleshoot a missing answer. If its preview
differs from the source, report the extraction problem.