> ## Documentation Index > Fetch the complete documentation index at: https://docs.useparagon.com/llms.txt > Use this file to discover all available pages before exploring further. # Get Result Rows > Filter and page through records in a stored tool result. Reads a `records`-shaped result. You can use the following query parameters to filter, trim, and page through the records: | Query | Meaning | | - | - | | `pointer` | JSON Pointer (`/` prefix) into the stored value. Defaults to the envelope `recordsPointer`. Depth is at most 8. | | `select` | Comma-separated field names to return for each record, at most 32. | | `where` | Repeatable filter clause in the format `field:operator:value`, at most 8. See [Filtering with `where`](#filtering-with-where). | | `pageSize` | Integer from 1 to 100. Default 50. | | `continuationToken` | `paging.continuationToken` from the previous page. | ```javascript theme={null} const params = new URLSearchParams({ select: "id,name", pageSize: "50", }); params.append("where", "status:eq:open"); const response = await fetch( `https://agentos.useparagon.com/api/tools/projects/${projectId}/results/${resultId}/rows?${params}`, { headers: { Authorization: `Bearer ${paragonUserToken}` } }, ); ``` ## Filtering with `where` `where` does not use a query language like SQL or OData. Each filter is a single colon-separated string, and you pass one `where` query parameter per filter. ### Clause format ```text theme={null} field:operator:value ``` | Part | Description | | - | - | | `field` | The name of a top-level key on each record, such as `status` or `Amount`. Field names are case-sensitive. Nested paths like `owner.name` are not supported, and a field name cannot contain a `:`. | | `operator` | One of the [operators](#operators) below. Operators are lowercase and case-sensitive. | | `value` | The value to compare against. Always written as text, with no quotes. Everything after the second `:` is the value, so values can contain colons, such as `createdAt:gte:2026-01-01T00:00:00Z`. Omit it for `exists`. | Values are not trimmed, quoted, or split on commas, so `name:eq: Acme` matches a leading space. Remember to URL-encode each clause when you build the query string. `URLSearchParams` does this for you. ### Combining filters Repeat the `where` parameter to add more filters. A record is returned only if it matches **every** clause (logical AND). There is no OR or NOT, and you can use at most 8 clauses per request. ```text theme={null} GET .../rows?where=status:eq:open&where=amount:gte:1000&where=name:contains:Acme ``` To express an OR, make one request for each alternative and merge the results in your agent. ### Operators | Operator | Matches when | Needs a value | | - | - | - | | `eq` | The field equals the value. | Yes | | `ne` | The field does not equal the value. Records that are missing the field also match. | Yes | | `gt` | The field is greater than the value. | Yes | | `gte` | The field is greater than or equal to the value. | Yes | | `lt` | The field is less than the value. | Yes | | `lte` | The field is less than or equal to the value. | Yes | | `contains` | The field, read as text, contains the value as a case-sensitive substring. | Yes | | `exists` | The field is present and is not `null`. | No | ### How values are compared Because every value in the query string is text, the comparison depends on what the field and the value look like: * **Numbers.** If a field is a number, or a string that parses as a number (like `"10"` or `"1e3"`), and the value also parses as a number, they are compared numerically. `Amount:eq:10.0` matches `10`, and `Amount:gt:9` matches `"10"`. Numeric strings are converted to numbers, so identifiers like `"00123"` and `"123"` are treated as equal. * **Strings.** If either side is not numeric, `eq` and `ne` compare the text exactly (case-sensitive). `gt`, `gte`, `lt`, and `lte` compare strings alphabetically when both sides are non-numeric. This works for ISO 8601 timestamps in the same format and time zone, such as `createdAt:gte:2026-01-01T00:00:00Z`, but it does not work for formats that don't sort alphabetically, like `01/02/2026`. * **Booleans.** Compare against the text `true` or `false`, such as `archived:eq:false`. Ordering operators never match booleans. * **Empty and missing values.** `gt`, `gte`, `lt`, and `lte` never match when the field is missing, `null`, empty, a boolean, an object, or an array, or when the value is empty or can't be compared to the field (such as `Amount:gt:abc` against a numeric field). `contains` never matches a missing or `null` field, or an empty value. `eq` with an empty value (`name:eq:`) only matches fields that are an empty string. * **Objects and arrays.** Comparison operators are meant for scalar fields. Use `exists` to check that a nested value is present, and use `select` to read it. ### Filtering and selecting together Filters run before `select` and paging, so you can filter on a field you do not return. `select` only controls which fields appear in each record in `data`; fields that are not on a record are omitted. ```javascript theme={null} const params = new URLSearchParams({ select: "id,name", pageSize: "25" }); params.append("where", "status:eq:open"); params.append("where", "amount:gte:1000"); params.append("where", "closedAt:exists"); ``` ### Paging through filtered results `matched` is the total number of records that match your filters across all pages, not only the page returned. When `paging.continuationToken` is not `null`, there are more matches. Pass it as `continuationToken` on the next request. A continuation token is tied to the `pointer`, `select`, and `where` clauses (in the same order) of the request that created it. If you change any of them, the request fails with `RESULT_SLICE_INVALID`. Start a new query without a token when you want different filters. ### Errors An invalid request returns `400` with the error code `RESULT_SLICE_INVALID` and a message describing the problem. Common causes: | Message | Cause | | - | - | | `Invalid where clause "..."` | The clause is missing a field or an operator, such as `status` or `:eq:open`. | | `Unknown filter operator "..."` | The operator is not one of the eight listed above, or is not lowercase. | | `where exceeds 8 clauses.` | More than 8 `where` parameters were sent. | | `The continuation token is invalid.` | The token was altered, or was created with a different `pointer`, `select`, or `where`. | ## OpenAPI ````yaml agent-os/openapi/tools.json GET /api/tools/projects/{projectId}/results/{resultId}/rows openapi: 3.1.0 info: title: Agent OS Tools API version: 1.0.0 description: Enable your agent to take action in 130+ different apps with Agent OS Tools. servers: - url: https://agentos.useparagon.com description: Production US Cloud - url: https://agentos.eu.paragon.so description: Production EU Cloud security: - bearerAuth: [] paths: /api/tools/projects/{projectId}/results/{resultId}/rows: get: summary: Get Result Rows operationId: getResultRows parameters: - $ref: '#/components/parameters/ProjectId' - name: resultId in: path required: true schema: type: string - name: pointer in: query schema: type: string description: RFC 6901 JSON Pointer. Must start with /. - name: select in: query schema: type: string description: Comma-separated fields, at most 32. - name: where in: query schema: type: array items: type: string style: form explode: true description: >- Repeatable filter in the format field:operator:value, at most 8. Clauses are combined with AND. field is a top-level record key. Operators (lowercase): eq, ne, gt, gte, lt, lte, contains, exists. Everything after the second colon is the value, and exists takes no value. Example: status:eq:open. - name: pageSize in: query schema: type: integer minimum: 1 maximum: 100 default: 50 - name: continuationToken in: query schema: type: string responses: '200': description: One page of records. content: application/json: schema: type: object required: - data - paging - matched - tokenBudget properties: data: type: array items: type: object additionalProperties: true paging: type: object required: - pageSize - continuationToken properties: pageSize: type: integer continuationToken: oneOf: - type: string - type: 'null' matched: type: integer tokenBudget: type: object required: - fullOutputTokens - returnedTokens - savedTokens properties: fullOutputTokens: type: integer returnedTokens: type: integer savedTokens: type: integer example: data: - id: '1' name: Open issue paging: pageSize: 50 continuationToken: null matched: 1 tokenBudget: fullOutputTokens: 20 returnedTokens: 8 savedTokens: 12 '400': $ref: '#/components/responses/Error' '401': $ref: '#/components/responses/Error' '404': $ref: '#/components/responses/Error' components: parameters: ProjectId: name: projectId in: path required: true schema: type: string format: uuid responses: Error: description: Error envelope. content: application/json: schema: $ref: '#/components/schemas/ErrorEnvelope' example: error: code: REQUEST_INVALID message: The request payload failed validation. hint: Correct the fields named in details and retry. retryable: false requestId: req_01hxyz details: - field: query issue: query should not be empty schemas: ErrorEnvelope: type: object required: - error properties: error: type: object required: - code - message - hint - retryable - requestId properties: code: type: string message: type: string hint: type: string retryable: type: boolean retryAfterMs: type: integer requestId: type: string details: type: array items: type: object required: - field - issue properties: field: type: string issue: type: string securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: JWT description: >- Your [Paragon User Token](/getting-started/installing-the-connect-sdk#setup) (JWT), which you can generate using your project's signing keys. ````