> ## Documentation Index > Fetch the complete documentation index at: https://developers.askparable.com/llms.txt > Use this file to discover all available pages before exploring further. # Read which AI processors may use Workspace data > Returns the selected Workspace’s AI model policy: the processors Parable allows for it, the processors the Workspace switched off, its default model per capability, and the effective policy every AI call applies. Requires permission to manage the Workspace’s AI processors. ## Access An MCP token or a Parable token can call this operation. See [Authentication](/getting-started/authentication.md) for the kinds of token. Your roles must grant `tenant.ai.manage`. ## Errors * `WA-TN-001`: The request has no Workspace context. Select a Workspace before reading its model policy. ## OpenAPI ````yaml /api-reference/openapi.json get /api/model-policy openapi: 3.0.3 info: description: >- Use the Workspace API to work with artifacts, connectors, Parables, and account resources. title: Workspace API version: v2026.10.22.1 servers: [] security: [] paths: /api/model-policy: get: tags: - Model policy summary: Read which AI processors may use Workspace data description: >- Returns the selected Workspace’s AI model policy: the processors Parable allows for it, the processors the Workspace switched off, its default model per capability, and the effective policy every AI call applies. Requires permission to manage the Workspace’s AI processors. operationId: ModelPolicyGetModelPolicyHandler parameters: - description: >- Workspace slug. Overrides subdomain-based workspace resolution when provided. in: header name: X-Tenant required: false schema: example: acme type: string - description: Optional request correlation ID echoed back in the response. in: header name: X-Request-ID required: false schema: example: 550e8400-e29b-41d4-a716-446655440000 format: uuid type: string responses: '200': content: application/json: schema: properties: data: $ref: '#/components/schemas/WorkspaceModelPolicy' links: $ref: '#/components/schemas/ResponseLinks' meta: $ref: '#/components/schemas/ResponseMeta' required: - data - meta type: object description: Successful response headers: X-Request-ID: $ref: '#/components/headers/X-Request-ID' '400': content: application/problem+json: schema: $ref: '#/components/schemas/ProblemDetails' description: The request is not valid. `detail` says why. headers: X-Request-ID: $ref: '#/components/headers/X-Request-ID' '401': content: application/problem+json: schema: $ref: '#/components/schemas/ProblemDetails' description: The token is missing, malformed, revoked, or expired. headers: X-Request-ID: $ref: '#/components/headers/X-Request-ID' '403': content: application/problem+json: schema: $ref: '#/components/schemas/ProblemDetails' description: >- The token cannot make this request. For example, the roles of its owner do not allow the operation. headers: X-Request-ID: $ref: '#/components/headers/X-Request-ID' '429': content: text/plain: schema: example: Too Many Requests type: string description: >- More than 120 requests a minute came from your address. Wait for the number of seconds in `Retry-After`, then try again. headers: Retry-After: description: The number of seconds to wait before you try again. schema: type: integer X-Request-ID: $ref: '#/components/headers/X-Request-ID' '500': content: application/problem+json: schema: $ref: '#/components/schemas/ProblemDetails' description: The request failed on the server. headers: X-Request-ID: $ref: '#/components/headers/X-Request-ID' security: - bearerAuth: [] x-codeSamples: - label: TypeScript SDK lang: typescript source: |- import { WebApiSDK } from "@parable-platform/web-api-sdk"; const sdk = new WebApiSDK({ baseUrl: process.env.PARABLE_API_URL!, auth: { token: process.env.PARABLE_API_TOKEN! }, requestInterceptor: (config) => ({ ...config, headers: { ...config.headers, "X-Tenant": process.env.PARABLE_WORKSPACE! }, }), }); const result = await sdk.modelPolicy.getModelPolicy(); - label: Python SDK lang: python source: |- import os from parable_web_api_sdk import ClientConfig, WebApiSDK sdk = WebApiSDK( ClientConfig( base_url=os.environ["PARABLE_API_URL"], auth_token=os.environ["PARABLE_API_TOKEN"], default_headers={"X-Tenant": os.environ["PARABLE_WORKSPACE"]}, ) ) result = sdk.model_policy.get_model_policy() - label: Rust SDK lang: rust source: >- use parable_web_api_sdk::{ClientConfig, RequestOptions, WebApiSdk}; let sdk = WebApiSdk::new(ClientConfig::with_base_url( std::env::var("PARABLE_API_URL")?, Some(std::env::var("PARABLE_API_TOKEN")?), None, ))?; let options = RequestOptions::with_header("X-Tenant", std::env::var("PARABLE_WORKSPACE")?, None, None); let result = sdk.model_policy.get_model_policy( Some(&options) ).await?; - label: Go SDK lang: go source: "sdk, err := webapisdk.New(webapisdk.SDKConfig{\n\tBaseURL: os.Getenv(\"PARABLE_API_URL\"),\n\tAuth: &webapisdk.AuthConfig{Token: os.Getenv(\"PARABLE_API_TOKEN\")},\n\tRequestInterceptor: func(ctx context.Context, req *http.Request) error {\n\t\treq.Header.Set(\"X-Tenant\", os.Getenv(\"PARABLE_WORKSPACE\"))\n\t\treturn nil\n\t},\n})\nif err != nil {\n\treturn err\n}\n\nresult, err := sdk.ModelPolicyNamespace.GetModelPolicy(ctx)" components: schemas: WorkspaceModelPolicy: description: >- A Workspace's AI model policy as its administrators see it: the stored choices and the effective policy Parable computes from them. Parable sets the processors the Workspace may use and its Vertex AI project. The Workspace switches processors off and sets the defaults. properties: capabilityDefaults: description: The Workspace's default model per capability. items: $ref: '#/components/schemas/ModelCapabilityDefault' type: array disabledProcessors: description: Processors the Workspace switched off. items: $ref: '#/components/schemas/ModelProcessor' type: array effective: $ref: '#/components/schemas/EffectiveModelPolicy' description: What every AI call over the Workspace's data applies. eligibleProcessors: description: >- The processors Parable allows for the Workspace. Null means Parable set no limit. An empty list means no processor may receive the Workspace's data. items: $ref: '#/components/schemas/ModelProcessor' nullable: true type: array features: description: >- The product feature each capability serves, from the environment policy. items: $ref: '#/components/schemas/ModelCapabilityFeature' type: array processors: description: >- Every processor the environment lists for any capability, in processor order, with its terms and what it serves for the Workspace. items: $ref: '#/components/schemas/WorkspaceModelProcessor' type: array settableDefaults: description: >- The capabilities the Workspace may set a default for, in capability order, with the kind of choice each takes. setModelPreferences refuses a default for any other capability. items: $ref: '#/components/schemas/WorkspaceSettableDefault' type: array vertexProject: description: >- The Workspace's own Vertex AI project, which Parable owns and sets. Null means the environment's shared project. nullable: true type: string required: - disabledProcessors - capabilityDefaults - effective - processors - features - settableDefaults type: object ResponseLinks: properties: first: nullable: true type: string last: nullable: true type: string next: nullable: true type: string prev: nullable: true type: string type: object ResponseMeta: properties: requestId: description: Unique identifier for the request type: string totalCount: description: Total count of items for paginated responses nullable: true type: integer required: - requestId type: object ProblemDetails: description: An error response in the Problem Details format (RFC 9457). properties: code: description: The stable error code, when the problem has one. example: WA-VL-001 type: string detail: description: What went wrong with this request. type: string details: description: >- Structured facts about the refusal, such as the entity or the path it concerns. Only some refusals have it. errors: description: >- The fields or filters that failed validation. Only a validation or filter error has it. requestId: description: >- The ID of the request, when the problem has one. The `X-Request-ID` response header always has it. type: string status: description: The HTTP status code. example: 400 type: integer title: description: A short summary, usually the text of the HTTP status. example: Bad Request type: string type: description: >- The problem type. It is `about:blank`: the status and `code` identify the problem. example: about:blank type: string required: - type - title - status - detail type: object ModelCapabilityDefault: additionalProperties: false description: A Workspace's default model for one capability. properties: capability: $ref: '#/components/schemas/ModelCapability' choice: $ref: '#/components/schemas/ModelChoice' required: - capability - choice type: object ModelProcessor: description: >- An AI processor that may receive a Workspace's data when Parable runs a model over it. The Workspace model policy is a set of these: Parable sets the processors a Workspace may use, the Workspace may switch some of them off, and the environment policy caps both. enum: - VERTEX - OPENROUTER_ZDR - ANTHROPIC - TYPESAFE type: string EffectiveModelPolicy: additionalProperties: false description: >- A Workspace's effective model policy: what Parable computes from the environment policy, the Workspace's stored model policy and the global model catalog. Every AI call over the Workspace's data applies one rule to it. properties: capabilities: items: $ref: '#/components/schemas/EffectiveModelCapability' type: array digest: description: |- SHA-256 of everything else in this policy. It changes whenever the environment policy, the stored policy or a catalog route it names changes, so consumers cache by it. maxLength: 64 minLength: 64 pattern: ^[0-9a-f]{64}$ type: string policyVersion: description: >- The stored policy's version (tenant_model_policy._version). 0 when the Workspace has no stored policy. format: int64 maximum: 9007199254740991 minimum: 0 type: integer tenantId: description: UUID v4 with automatic base62 encoding for client-facing APIs pattern: >- ^([0-9A-Za-z]{1,22}|[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12})$ type: string vertexProject: description: >- The Workspace's own Vertex AI project, which Parable owns, once Parable has set it up. Null means the environment's shared project: none is set, or the one set is not ready yet. Vertex calls for the Workspace run in this project. It is routing only and never part of a model identity. nullable: true type: string required: - tenantId - policyVersion - digest - capabilities type: object ModelCapabilityFeature: additionalProperties: false description: |- The product feature one capability serves, and what a Workspace loses when no processor may serve it. The Workspace AI processors page shows it before a narrowing or a default is saved. properties: capability: $ref: '#/components/schemas/ModelCapability' name: description: The feature's name, as a Workspace administrator reads it. type: string whenDefaultChanges: description: |- What a change to the Workspace's default model does to work already scheduled. Null when a change reaches only later calls. nullable: true type: string whenUnavailable: description: What stops working when no processor may serve the capability. type: string required: - capability - name - whenUnavailable type: object WorkspaceModelProcessor: description: >- One AI processor as a Workspace's administrators see it: what it is, its terms, and what it serves for the Workspace. Parable computes every field. properties: capabilities: description: >- The capabilities this environment lets the processor serve, when Parable allows the processor for the Workspace, whether or not the Workspace switched it off. Empty when Parable does not allow it. items: $ref: '#/components/schemas/ModelCapability' type: array eligibleCapabilities: description: >- The capabilities it serves for the Workspace now: the ones above, unless the Workspace switched it off. items: $ref: '#/components/schemas/ModelCapability' type: array ponderServer: description: |- The ponder() server that sends a call to this processor, for a SQL_MODEL default. Null when ponder() cannot reach it. nullable: true type: string processor: $ref: '#/components/schemas/ModelProcessor' switchedOff: description: Whether the Workspace switched it off. type: boolean terms: description: |- What the processor is and its retention terms, from the environment policy. Null when the environment records none for it. oneOf: - $ref: '#/components/schemas/ModelProcessorTerms' - enum: - null nullable: true type: object withinCeiling: description: >- Whether Parable allows the processor for the Workspace. The Workspace can switch only an allowed processor off, and back on. type: boolean required: - processor - capabilities - eligibleCapabilities - withinCeiling - switchedOff type: object WorkspaceSettableDefault: description: |- A capability a Workspace may set its own default model for, and the one kind of choice that default takes. Embedding models are platform-only, so they never appear Decision 6). properties: capability: $ref: '#/components/schemas/ModelCapability' choiceKind: $ref: '#/components/schemas/ModelChoiceKind' required: - capability - choiceKind type: object ModelCapability: description: |- A kind of model call that runs over a Workspace's data. The environment policy lists the processors each one may use, and a Workspace sets a default model for the ones it may choose. Parable chooses the models of the embedding capabilities: a Workspace sets no default for them. enum: - PONDER_CHAT - SQL_EMBED - SQL_MODEL - SQL_JEV - POOL_SEARCH_EMBEDDING - SCHEMA_INFERENCE - PONDER_CLEANUP - PONDER_NAMING - PONDER_JUDGMENT - ROW_SEARCH_EMBEDDING - PONDER_LANGUAGE type: string ModelChoice: description: >- The model one capability uses, either a catalog Model Release, an embedding model identity, a Jev model version, a processor's own model id, or ponder() model coordinates. discriminator: mapping: EMBED_MODEL: '#/components/schemas/EmbedModelChoice' JEV_MODEL: '#/components/schemas/JevModelChoice' MODEL_RELEASE: '#/components/schemas/ModelReleaseChoice' PONDER_MODEL: '#/components/schemas/PonderModelChoice' PROCESSOR_MODEL: '#/components/schemas/ProcessorModelChoice' propertyName: kind oneOf: - $ref: '#/components/schemas/ModelReleaseChoice' - $ref: '#/components/schemas/EmbedModelChoice' - $ref: '#/components/schemas/JevModelChoice' - $ref: '#/components/schemas/ProcessorModelChoice' - $ref: '#/components/schemas/PonderModelChoice' EffectiveModelCapability: additionalProperties: false description: One capability of a Workspace's effective model policy. properties: capability: $ref: '#/components/schemas/ModelCapability' eligibleProcessors: description: >- The environment's processors for the capability that Parable allows for the Workspace, minus the ones the Workspace switched off, in environment order. items: $ref: '#/components/schemas/ModelProcessor' type: array refusal: description: >- Set when a call that names no model must refuse. Null when resolved is set. oneOf: - $ref: '#/components/schemas/ModelPolicyRefusal' - enum: - null nullable: true type: string resolved: description: >- The model a call that names none uses: the Workspace default when it is set and eligible, else the first platform default candidate whose processor is eligible. oneOf: - $ref: '#/components/schemas/ResolvedModelChoice' - enum: - null nullable: true type: object required: - capability - eligibleProcessors type: object ModelProcessorTerms: additionalProperties: false description: |- What one AI processor is and the terms under which it receives a Workspace's data. The Workspace AI processors page shows it, and it is the one source for the AI entries of the sub-processor list. properties: description: description: What the processor is and how Parable reaches it. type: string name: description: The processor's name, as a Workspace administrator reads it. type: string processor: $ref: '#/components/schemas/ModelProcessor' retention: description: The retention terms that apply to the Workspace data it receives. type: string retentionBasis: $ref: '#/components/schemas/ModelProcessorRetentionBasis' required: - processor - name - description - retention - retentionBasis type: object ModelChoiceKind: description: Which member of ModelChoice a value is. enum: - MODEL_RELEASE - EMBED_MODEL - JEV_MODEL - PROCESSOR_MODEL - PONDER_MODEL type: string EmbedModelChoice: additionalProperties: false description: |- An embedding model. Parable chooses the embedding models: a Workspace cannot. As a capability's defaultChoice or a Workspace default the processor is VERTEX; a platform candidate (PlatformModelCandidate) names its own processor. properties: identity: $ref: '#/components/schemas/EmbedModelIdentity' kind: $ref: '#/components/schemas/ModelChoiceKind' required: - kind - identity type: object JevModelChoice: additionalProperties: false description: >- A Jev model version, such as `jev-1.13.0`. TypeSafe serves Jev, so the processor is TYPESAFE. properties: kind: $ref: '#/components/schemas/ModelChoiceKind' version: maxLength: 64 minLength: 1 type: string required: - kind - version type: object ModelReleaseChoice: additionalProperties: false description: A Model Release from the global catalog, for chat. properties: kind: $ref: '#/components/schemas/ModelChoiceKind' modelReleaseId: description: UUID v4 with automatic base62 encoding for client-facing APIs pattern: >- ^([0-9A-Za-z]{1,22}|[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12})$ type: string required: - kind - modelReleaseId type: object PonderModelChoice: additionalProperties: false description: >- The coordinates of a ponder() or model() call, for SQL_MODEL. The processor is the one the coordinates' server names: `vertex` is VERTEX, `anthropic` is ANTHROPIC, and `openrouter` is OPENROUTER_ZDR. properties: kind: $ref: '#/components/schemas/ModelChoiceKind' model: $ref: '#/components/schemas/PonderModelConfig' processor: $ref: '#/components/schemas/ModelProcessor' required: - kind - model - processor type: object ProcessorModelChoice: additionalProperties: false description: >- A model named by the processor that serves it, in that processor's own model id: `anthropic/claude-haiku-5-5` on OPENROUTER_ZDR, or `gemini-3.5-flash-lite` on VERTEX. The platform defaults use it because a catalog Model Release id differs between environments. properties: kind: $ref: '#/components/schemas/ModelChoiceKind' modelId: maxLength: 200 minLength: 1 type: string processor: $ref: '#/components/schemas/ModelProcessor' required: - kind - processor - modelId type: object ModelPolicyRefusal: description: >- Why a model call fails closed. A call never falls back to another processor. enum: - ENVIRONMENT_POLICY_MISSING - NO_ELIGIBLE_PROCESSOR - NO_MODEL - PROCESSOR_NOT_ELIGIBLE - MODEL_NOT_ROUTABLE type: string ResolvedModelChoice: additionalProperties: false description: >- A model a call resolved to, the processor that serves it, and where it came from. properties: choice: $ref: '#/components/schemas/ModelChoice' inputTokenLimit: description: >- The model's input limit in tokens, from the platform candidate that resolved (PlatformModelCandidate.inputTokenLimit). Null for any other choice. format: int64 maximum: 9007199254740991 minimum: -9007199254740991 nullable: true type: integer modelReleaseId: description: >- The catalog Model Release, for a chat choice: the release itself, or the release the catalog routes a processor model id to. nullable: true pattern: >- ^([0-9A-Za-z]{1,22}|[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12})$ type: string processor: $ref: '#/components/schemas/ModelProcessor' routeModelId: description: >- The processor's model id for a catalog Model Release, from its route. nullable: true type: string source: $ref: '#/components/schemas/ModelChoiceSource' required: - source - choice - processor type: object ModelProcessorRetentionBasis: description: >- On what basis a processor keeps, or does not keep, the Workspace data it receives. enum: - ZERO_DATA_RETENTION_REQUIRED - SERVICE_TERMS - NO_ZERO_DATA_RETENTION_AGREEMENT type: string EmbedModelIdentity: additionalProperties: false description: >- The identity that makes stored vectors comparable: two vectors compare only when every field matches. The server that computes a vector is routing, never identity, so it is absent. properties: dimensions: description: >- Signed 64-bit integer; range bounded by JavaScript's safe-integer ceiling. format: int64 maximum: 9007199254740991 minimum: 1 type: integer name: minLength: 1 type: string normalization: minLength: 1 type: string provider: minLength: 1 type: string variant: minLength: 1 type: string version: minLength: 1 type: string required: - provider - name - variant - version - dimensions - normalization type: object PonderModelConfig: description: |- Model coordinates and per-call inference settings packed by the ponder() and model() SQL functions. The model policy's SQL_MODEL choice (PonderModelChoice) carries the same coordinates. properties: date: type: string effort: type: string name: type: string provider: type: string request_timeout_secs: description: >- Signed integer count of seconds: an amount of elapsed time, never a point in time. An epoch timestamp is Temporal.DateTime with x-temporal-format: unix. format: int64 maximum: 9007199254740991 minimum: 0 nullable: true type: integer server: type: string tokens: description: >- Signed 64-bit integer; range bounded by JavaScript's safe-integer ceiling. format: int64 maximum: 9007199254740991 minimum: -9007199254740991 nullable: true type: integer variant: type: string version: type: string required: - provider - name - variant - version - date - effort - server type: object ModelChoiceSource: description: Where a resolved model came from. The members are in precedence order. enum: - EXPLICIT - REQUEST_OPTION - WORKSPACE_DEFAULT - PLATFORM_DEFAULT type: string headers: X-Request-ID: description: >- The ID of the request. It is the `X-Request-ID` value that you sent, or an ID that the server made. Give it when you report a problem. schema: type: string securitySchemes: bearerAuth: description: >- A personal access token, which starts with `pat_`. A Parable token can call each operation that the roles of its owner allow. An MCP token can call only the read operations that are open to agents. A SQL token does not work on this API. scheme: bearer type: http ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.