---
title: "Search Knowledge documents"
method: POST
path: "/api/agent-sources/search"
tags: ["agent_sources"]
---

# Search Knowledge documents

`POST /api/agent-sources/search`

## Request

- Base URL: `https://beta.getsales.io/unibox/v1`
- URL: `https://beta.getsales.io/unibox/v1/api/agent-sources/search`
- Auth: HTTP bearer and API key in header `Team-SID`

The AI SDR's Knowledge documents: the files and websites its answers were read from, newest first, each with its instruction (how it is read), where its read stands (reading with page n of m, read, failed with why) and when it was last read; counts per kind and status. include usage adds each document's answers and the replies that read one in 30 days; include parse its newest parse record. page_size 0 returns counts alone.

Contract:
- MCP tool `search_agent_sources`, registry package `mcp.unibox/agent_sources`, mount `unibox.knowledge`.
- Operation `search`, response envelope `search`.
- Flags: read only.

## Request body

- SearchAgentSourcesRequest: Request body of `search_agent_sources`.
  - `filter` object: Field filters, ANDed: {field: {eq | ne | in | nin | gte | lte | gt | lt | is_null}}; q inside it is free text. Omit for every row.
    - `q` string: Words in a document's title.
    - `sid` object: Document sid(s), ub_as_ prefix.
      - `eq` string: Knowledge document sid (ub_as_…).
      - `in` string[]
    - `agent_sid` object: The AI SDR the rows belong to (ub_ag_…, block 10, 2026-10-30); left out, the team's default agent.
      - `eq` string
      - `in` string[]
    - `kind` object: file or website.
      - `eq` 'file' | 'website'
      - `in` string[]
    - `status` object: { eq: "failed" } lists what the parse gave up on; { eq: "reading" } what is being read.
      - `eq` 'reading' | 'read' | 'failed'
      - `in` string[]
    - `created_at` object: ISO 8601 range.
      - `gte` string
      - `lte` string
      - `gt` string
      - `lt` string
  - `include` string[]: Relations to attach to every item; omit for the bare rows.
  - `sort` object: One column and its direction; omit for the default order.
    - `field` 'created_at' | 'title' | 'read_at', required: The column to order by.
    - `direction` 'asc' | 'desc': asc or desc; default desc.
  - `page_size` integer: 0..200, default 50; 0 = counts only, 1 = the first row.
  - `cursor` string, nullable: pagination.next_cursor of the previous page.

## Response `200`

`search` success envelope.

- SearchAgentSourcesResponse
  - `success` true, required
  - `operation` 'search', required
  - `items` object[], required
    - `item` object, required
      - `sid` string, required: Knowledge document sid (ub_as_…).
      - `team_sid` string, required: Owning team (ts_tm_…, external - gs.service.id).
      - `agent_sid` string, required: The team's agent (ub_ag_…).
      - `kind` 'file' | 'website', required: file: a file the team uploaded; website: a start address and the pages under it (a help center is a website).
      - `title` string, required: The name the list shows: a file's name until renamed, a website's host and path.
      - `status` 'reading' | 'read' | 'failed', required: reading while a parse runs (progress says page n of m), read once a parse finished (its answers wait for the OK or are in Knowledge), failed with the reason in failure.
      - `instruction` string, nullable, required: How the parse reads it, in the person's words, up to 600 chars: what to read out, what to skip, how exact. Suggested from the first pages (suggest_agent_source_instruction), edited by the person; null until one is written.
      - `file` object, nullable, required: A file's file; null for a website.
        - `name` string, required: The file's name as uploaded.
        - `format` 'pdf' | 'docx' | 'txt' | 'md', required
        - `mime` string, required: Its media type as uploaded.
        - `bytes` integer, required: Its size, 20 MB at most.
        - `pages` integer, nullable, required: Its pages (a PDF's, a Word file's as Word counted them); null for a text file.
      - `website` object, nullable, required: A website's address, scope, refresh and pages; null for a file.
        - `url` string, required: The start address.
        - `scope` string, required: The path under which pages are read; / for the whole site.
        - `refresh` 'weekly' | 'daily', required: How often it is read again by itself.
        - `pages` integer, nullable, required: How many pages the last read took; null until read.
      - `progress` object, required: How far the parse got: page n of m ("Reading, page 9 of 22"); 0 of 0 before the pages are known.
        - `done` integer, required
        - `total` integer, required
      - `failure` object, nullable, required: Why the parse gave up; null unless failed.
        - `reason` 'unreadable' | 'no_text' | 'too_large' | 'website_unreachable' | 'nothing_found' | 'model_failed', required
        - `message` string, required: What to do, in the words the list shows.
      - `read_at` string, nullable, required: ISO 8601 UTC of the last parse that finished; null until then.
      - `next_refresh_at` string, nullable, required: A website: ISO 8601 UTC of its next read by itself; null for a file.
      - `created_by` object, required: Who added it.
        - `actor_type` 'user' | 'support' | 'api_key' | 'system' | 'agent', required
        - `actor_sid` string, nullable, required
        - `team_sid` string, required
        - `actor_name` string, nullable: The OAuth client that acted ("Claude", "n8n"); null for a user, an API key or a system job.
        - `oauth_client_sid` string, nullable: The acting OAuth client (id_oc_*); null off the OAuth path.
        - `reason` string, nullable: Why a system actor wrote the row (snapshot_capture_job, ...); null otherwise.
        - `permissions` object, required: Internal audit context: the grant set the actor held at write time. Not a contract.
        - `cluster_id` integer, nullable: Internal audit context: the cluster that served the write.
        - `trace_id` string, nullable: Internal audit context: the trace id of the request that wrote the row.
      - `created_at` string, required: ISO 8601 UTC.
      - `updated_at` string, required: ISO 8601 UTC.
    - `included` object, required
      - `usage` object: include usage: its answers, and the replies that read one in 30 days.
        - `answers` integer, required: The answers read from it that Knowledge holds now.
        - `used` integer, required: The agent's replies that went out in the last 30 days whose turn read one of those answers.
      - `parse` object, nullable: include parse: its newest parse record without the answers it found (get_agent_source_parse reads them); null when it was never parsed.
        - `sid` string, required: Parse record sid (ub_ps_…).
        - `team_sid` string, required: Owning team (ts_tm_…, external - gs.service.id).
        - `agent_sid` string, required: The team's agent (ub_ag_…).
        - `source_sid` string, required: The document read (ub_as_…).
        - `kind` 'parse' | 'again' | 'refresh', required: parse: a first read, compared with every answer the agent has; again: Read it again or a replaced file, compared with the document's own answers first; refresh: a website's scheduled read, the same comparison.
        - `status` 'queued' | 'reading' | 'found' | 'applied' | 'failed' | 'discarded', required: queued for the job; reading (step says which); found (the answers wait for the OK); applied (what the person picked is in Knowledge; a read again that found everything as it was applies itself); failed (failure says why); discarded (closed with nothing added).
        - `instruction` string, nullable, required: The instruction this parse read by.
        - `step` 'reading' | 'splitting' | 'extracting' | 'comparing', nullable, required: Where a running parse stands: reading, splitting, extracting, comparing; null otherwise.
        - `steps` object, required: What it did, as the modal shows it; a step not reached is absent.
          - `pages` object: The pages read, skipped on the instruction's word, and in all.
            - `read` integer, required
            - `skipped` integer, required
            - `total` integer, required
          - `passages` integer: The passages split by heading.
          - `questions` integer: The questions found with their answers.
          - `compared` object: The rows per state.
            - `new` integer, required
            - `already_there` integer, required
            - `says_otherwise` integer, required
            - `unchanged` integer, required
            - `gone` integer, required
          - `applied` object: What apply did: answers added, written over, removed.
            - `added` integer, required
            - `updated` integer, required
            - `removed` integer, required
        - `found` object[], required: The answers found, each with its state against Knowledge; empty until the comparison finished (and on the include of search_agent_sources).
          - `id` integer, required: The row's id within the parse, from 1: what apply names.
          - `question` string, required: The question as a lead would ask it, up to 500 chars.
          - `phrasings` string[], required: Other ways a lead would ask it, up to twelve.
          - `answer` string, required: The answer in the document's words, up to 1200 chars.
          - `page` integer, nullable, required: The page it was read from (a website's page ordinal); null for a text file.
          - `url` string, nullable, required: A website page's address; null for a file.
          - `passage_from` integer, nullable, required: The passage's start offset in its page's text.
          - `passage_to` integer, nullable, required: The passage's end offset.
          - `passage` string, nullable, required: The passage it was read from, up to 1200 chars.
          - `state` 'new' | 'already_there' | 'says_otherwise' | 'unchanged' | 'gone', required: new: nothing in Knowledge answers it (comes checked); already_there: an answer with the same meaning exists (comes unchecked); says_otherwise: an answer exists and differs (Keep ours or Use the new one); unchanged: a read again found the document's own answer as it was (its sid and reads kept); gone: an earlier answer of the document the read again did not find (listed to remove).
          - `similarity` number, nullable, required: How near the existing answer's question is, (1 + cosine) / 2; null on a new row.
          - `existing` object, nullable, required: The existing answer the row was compared with; null on a new row.
            - `sid` string, required: The existing answer (ub_an_…).
            - `question` string, required: Its question.
            - `answer` string, required: Its answer.
            - `source_kind` string, required: document or manual.
            - `source_sid` string, nullable, required: Its document (ub_as_…) when read from one.
            - `source_title` string, nullable, required: That document's title.
            - `page` integer, nullable, required: That document's page.
        - `failure` object, nullable, required: Why the parse gave up; null unless failed.
          - `reason` 'unreadable' | 'no_text' | 'too_large' | 'website_unreachable' | 'nothing_found' | 'model_failed', required
          - `message` string, required: What to do, in the words the modal shows.
        - `read_at` string, nullable, required: ISO 8601 UTC when the comparison finished; null until then.
        - `applied_at` string, nullable, required: ISO 8601 UTC when the answers were taken in; null until then.
        - `applied_by` object, nullable, required: Who took them in; null on a read that applied itself.
          - `actor_type` 'user' | 'support' | 'api_key' | 'system' | 'agent', required
          - `actor_sid` string, nullable, required
          - `team_sid` string, required
          - `actor_name` string, nullable: The OAuth client that acted ("Claude", "n8n"); null for a user, an API key or a system job.
          - `oauth_client_sid` string, nullable: The acting OAuth client (id_oc_*); null off the OAuth path.
          - `reason` string, nullable: Why a system actor wrote the row (snapshot_capture_job, ...); null otherwise.
          - `permissions` object, required: Internal audit context: the grant set the actor held at write time. Not a contract.
          - `cluster_id` integer, nullable: Internal audit context: the cluster that served the write.
          - `trace_id` string, nullable: Internal audit context: the trace id of the request that wrote the row.
        - `created_by` object, required: Who opened the parse; the system on a website's refresh.
          - `actor_type` 'user' | 'support' | 'api_key' | 'system' | 'agent', required
          - `actor_sid` string, nullable, required
          - `team_sid` string, required
          - `actor_name` string, nullable: The OAuth client that acted ("Claude", "n8n"); null for a user, an API key or a system job.
          - `oauth_client_sid` string, nullable: The acting OAuth client (id_oc_*); null off the OAuth path.
          - `reason` string, nullable: Why a system actor wrote the row (snapshot_capture_job, ...); null otherwise.
          - `permissions` object, required: Internal audit context: the grant set the actor held at write time. Not a contract.
          - `cluster_id` integer, nullable: Internal audit context: the cluster that served the write.
          - `trace_id` string, nullable: Internal audit context: the trace id of the request that wrote the row.
        - `created_at` string, required: ISO 8601 UTC.
        - `updated_at` string, required: ISO 8601 UTC.
  - `pagination` object, required
    - `next_cursor` string, nullable, required
    - `has_more` boolean, required
    - `total_count` integer, nullable, required
  - `applied_filters` object, required
  - `includes` string[], required
  - `meta` object, required
    - `trace_id` string, required: UUID v7; same 128-bit value as the X-Trace-Id header.
    - `span_id` string, required: 16 hex chars, root span of this request.
    - `timestamp` string, required: ISO 8601 UTC (Y-m-dTH:i:sZ), response time.
    - `duration_ms` integer, required: Server-side wall clock.
    - `team_sid` string, nullable: The team this call ran in (the token team, or the team_sid override). Null when unauthenticated; absent from pre-2026-08-20 backends.
    - `actor_type` string, nullable: user | agent | api_key | system. Null when unauthenticated; absent from pre-2026-08-20 backends.
  - `counts` object
    - `total_count` integer, required: Documents matching the filter (full filter scope, not the page).
    - `groups` object, required
      - `kind` union, required: Documents by kind (file, website) over the other filters.
        - object
        - unknown[]
          - unknown
      - `status` union, required: Documents by status (reading, read, failed) over the other filters.
        - object
        - unknown[]
          - unknown
      - `answers` union, required: The agent's answers per source kind, whatever the filter: document (read from a document) and manual (added by hand): the block on top of Knowledge.
        - object
        - unknown[]
          - unknown

## Other responses

- `4XX`: MCP error envelope. `error.code` is one of validation_failed, nothing_to_update, not_found, relation_not_found, invalid_transition, limit_exceeded, payment_required, duplicate_rejected, conflict, delete_blocked, unauthorized, forbidden, rate_limited, not_implemented.
- `5XX`: MCP error envelope with `error.code` internal_error or service_unavailable.

---

[API](https://skmtc.dev/getsales/apis/getsales-public-contract-gs-service-unibox.md) - [All operations](https://skmtc.dev/getsales/apis/getsales-public-contract-gs-service-unibox/llms.txt) - [OpenAPI document](https://skmtc.dev/getsales/apis/getsales-public-contract-gs-service-unibox/revisions/8f1bc317897f?raw)