> ## Documentation Index > Fetch the complete documentation index at: https://docs.meibel.ai/llms.txt > Use this file to discover all available pages before exploring further. # Update Agent ## OpenAPI ````yaml /api-v2.json put /agents/{agent_id} openapi: 3.1.0 info: title: Meibel AI API summary: Meibel Gateway Service description: >- Our API allows you to interact with our services. Read the [docs](https://docs.meibel.ai) to learn how to use it. version: 0.6.4 servers: - url: https://api.meibel.ai/v2 description: Meibel API (v2) - url: https://api.dev.meibel.ai/v2 description: Meibel Dev API (v2) - url: http://localhost:8000/v2 description: Local Development security: [] tags: - name: Datasources description: v2 datasource management - name: Data Elements description: v2 data element CRUD and search - name: File Upload description: Upload files to a datasource and track upload progress - name: Ingest description: Trigger and monitor datasource ingestion - name: Datasource Downloads description: Export all datasource content as a zip archive - name: Table Descriptions description: v2 table and column descriptions - name: Metadata Model Catalog description: v2 metadata model catalog - name: Confidence Scoring description: v2 confidence scoring jobs and summaries - name: Documents description: Parse and transform documents into structured data - name: Agents description: Agent definition management - name: Sessions description: Agent session (execution) management and chat - name: Execution Policies description: Data access and tool usage constraints scoped to agent sessions - name: Artifact Schemas description: Agent artifact schema management paths: /agents/{agent_id}: put: tags: - Agents summary: Update Agent operationId: updateAgent parameters: - name: agent_id in: path required: true schema: type: string title: Agent Id requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/UpdateAgentDefinitionRequest' responses: '200': description: Successful Response content: application/json: schema: $ref: '#/components/schemas/UpdateAgentDefinitionResponse' '422': description: Validation Error content: application/json: schema: $ref: '#/components/schemas/HTTPValidationError' security: - APIKeyHeader: [] x-codeSamples: - lang: Python label: Python source: |- import os from meibel import MeibelClient from meibel.models import UpdateAgentDefinitionRequest client = MeibelClient(api_key=os.environ["MEIBEL_API_KEY"]) request = UpdateAgentDefinitionRequest( display_name="Customer Support Agent", instructions="You are a helpful customer support agent for Acme Corp.", llm_model="gpt-4o", temperature=0.3, max_tokens=2048, tags=["support", "production"], ) response = client.agents.update("agent_abc123", request) print(f"Updated agent {response.id}, version {response.version}") - lang: TypeScript label: TypeScript source: >- import { MeibelClient } from "meibel"; const client = new MeibelClient({ apiKey: process.env.MEIBEL_API_KEY }); const updatedAgent = await client.agents.update("agent_abc123", { display_name: "Support Assistant", instructions: "You are a helpful customer support agent for Acme Corp. Be concise and friendly.", llm_model: "gpt-4o", temperature: 0.3, max_tokens: 2048, tools: [ { name: "search_knowledge_base", type: "rag_search", description: "Search the support knowledge base for relevant articles", config: { datasource_id: "ds_kb_9f8e7d6c" }, use_for: ["answering product questions", "troubleshooting steps"], }, ], execution_policy: { datasources: { ds_kb_9f8e7d6c: { documents: { filter: { "data_element.__name__": { $nin: ["internal_notes.pdf"] } }, }, }, }, }, tags: ["support", "production"], }); console.log(`Updated agent ${updatedAgent.id} to version ${updatedAgent.version}`); console.log(`Catalog URN: ${updatedAgent.catalog_urn}`); - lang: Go label: Go source: "package main\n\nimport (\n\t\"context\"\n\t\"fmt\"\n\t\"os\"\n\n\tv2 \"github.com/meibel-ai/meibel-go/v2\"\n)\n\nfunc main() {\n\tclient := v2.NewClient(v2.WithAPIKey(os.Getenv(\"MEIBEL_API_KEY\")))\n\n\tresp, err := client.Agents.Update(context.Background(), \"agent_abc123\", v2.UpdateAgentDefinitionRequest{\n\t\tDisplayName: v2.String(\"Support Assistant\"),\n\t\tInstructions: v2.String(\"You are a helpful customer support agent for Acme Corp.\"),\n\t\tLlmModel: v2.String(\"gpt-4o\"),\n\t\tTemperature: v2.Float64(0.3),\n\t\tMaxTokens: v2.Int(2048),\n\t\tTools: []v2.AgentToolDefinition{\n\t\t\t{\n\t\t\t\tName: \"order_lookup\",\n\t\t\t\tType: \"database_query\",\n\t\t\t\tDescription: v2.String(\"Look up order details by order ID\"),\n\t\t\t\tConfig: map[string]any{\n\t\t\t\t\t\"datasource_id\": \"ds_orders_prod\",\n\t\t\t\t},\n\t\t\t},\n\t\t},\n\t\tTags: []string{\"support\", \"production\"},\n\t})\n\tif err != nil {\n\t\tpanic(err)\n\t}\n\n\tfmt.Printf(\"Updated agent %s to version %s\\n\", resp.ID, resp.Version)\n}" components: schemas: UpdateAgentDefinitionRequest: properties: display_name: anyOf: - type: string - type: 'null' title: Display Name description: Human-readable name of the agent instructions: anyOf: - type: string - type: 'null' title: Instructions description: System prompt/instructions type: anyOf: - type: string - type: 'null' title: Type description: Agent type description: anyOf: - type: string - type: 'null' title: Description description: Description of the agent llm_model: anyOf: - type: string - type: 'null' title: Llm Model description: LLM model to use fallback_models: anyOf: - items: type: string type: array - type: 'null' title: Fallback Models description: List of fallback models datasources: anyOf: - items: type: string type: array - type: 'null' title: Datasources description: Datasource IDs the agent has access to tools: anyOf: - items: $ref: '#/components/schemas/AgentToolDefinition' type: array - type: 'null' title: Tools description: Tools configuration artifacts: anyOf: - items: type: string type: array - type: 'null' title: Artifacts description: Catalog URNs of artifacts the agent produces confidence_configs: anyOf: - items: type: string type: array - type: 'null' title: Confidence Configs description: Confidence scoring module names to apply during execution temperature: anyOf: - type: number maximum: 2 minimum: 0 - type: integer maximum: 2 minimum: 0 - type: 'null' title: Temperature description: LLM temperature max_tokens: anyOf: - type: integer - type: 'null' title: Max Tokens description: Maximum tokens in response tags: anyOf: - items: type: string type: array - type: 'null' title: Tags description: Tags for categorization icon: anyOf: - type: string - type: 'null' title: Icon description: UI icon identifier execution_policy: anyOf: - $ref: '#/components/schemas/ExecutionPolicy' - type: 'null' description: Inline execution policy constraints (datasources, tools) execution_policy_ids: anyOf: - items: type: string type: array - type: 'null' title: Execution Policy Ids description: IDs of stored ExecutionPolicies to compose type: object title: UpdateAgentDefinitionRequest description: >- Request model for updating an agent definition. Name is intentionally excluded as it serves as the stable identifier for a version chain and cannot be changed. UpdateAgentDefinitionResponse: properties: id: type: string title: Id description: New agent definition ID catalog_urn: type: string title: Catalog Urn description: Catalog URN for the new version version: type: string title: Version description: New version number type: object required: - id - catalog_urn - version title: UpdateAgentDefinitionResponse description: Response model for updating an agent definition. HTTPValidationError: properties: detail: items: $ref: '#/components/schemas/ValidationError' type: array title: Detail type: object title: HTTPValidationError AgentToolDefinition: properties: name: type: string title: Name description: Instance name - what the LLM sees and calls type: type: string title: Type description: 'Tool type: rag_search, database_query, etc.' description: anyOf: - type: string - type: 'null' title: Description description: Description shown to LLM default: '' config: anyOf: - additionalProperties: true type: object - type: 'null' title: Config description: >- Tool config passed to activity via tool_context (datasource_id, base_prompt, etc.) parameters_schema: anyOf: - additionalProperties: true type: object - type: 'null' title: Parameters Schema description: Optional override for the tool's parameters schema use_for: anyOf: - items: type: string type: array - type: 'null' title: Use For description: When to use this tool (injected into system prompt) avoid_for: anyOf: - items: type: string type: array - type: 'null' title: Avoid For description: When NOT to use this tool (injected into system prompt) require_approval: anyOf: - type: boolean - type: 'null' title: Require Approval description: >- If true, workflow pauses for human approval before executing this tool default: false approval_message: anyOf: - type: string - type: 'null' title: Approval Message description: >- Message to display when requesting approval (supports {{variable}} templates) type: object required: - name - type title: AgentToolDefinition description: AgentToolDefinition ExecutionPolicy: properties: datasources: anyOf: - additionalProperties: $ref: '#/components/schemas/DatasourceView' type: object - type: 'null' title: Datasources description: >- Per-datasource access controls, keyed by datasource_id. Each entry can disable the datasource entirely or apply TAG/RAG filters to restrict which data is accessible. tools: anyOf: - additionalProperties: $ref: '#/components/schemas/ToolConfig' type: object - type: 'null' title: Tools description: >- Per-tool access controls, keyed by tool instance name. Each entry can disable the tool entirely or constrain its parameters to specific values or ranges. type: object title: ExecutionPolicy description: >- Session-level access constraints (hardrails) for data and tools. Controls what data an agent session can access and what tools it can use. Constraints are enforced by the platform at runtime — the agent cannot bypass them. Multiple policies can be composed: stored policies (by ID) and/or an inline policy are structurally merged. Overlapping datasource or tool entries are combined with ``$and`` (intersection semantics). All filter fields use MongoDB-style constraint operators. Plain values are automatically normalized to ``{"$eq": value}``. ValidationError: properties: loc: items: anyOf: - type: string - type: integer type: array title: Location msg: type: string title: Message type: type: string title: Error Type input: title: Input ctx: type: object title: Context type: object required: - loc - msg - type title: ValidationError DatasourceView: properties: disabled: anyOf: - type: boolean - type: 'null' title: Disabled description: >- When true, the datasource is entirely inaccessible for this session. All queries against it will be blocked. tables: anyOf: - $ref: '#/components/schemas/TagConstraints' - type: 'null' description: >- TAG (SQL/database) constraints. Controls table/row access and column redaction. documents: anyOf: - $ref: '#/components/schemas/RagConstraints' - type: 'null' description: >- RAG (vector search) constraints. Controls which documents can be retrieved. type: object title: DatasourceView description: >- Access controls for a single datasource. Each datasource can be entirely disabled, or selectively constrained via TAG (SQL) and/or RAG (vector search) filters. Keyed by datasource_id in ``ExecutionPolicy.datasources``. ToolConfig: properties: disabled: anyOf: - type: boolean - type: 'null' title: Disabled description: >- When true, the tool is removed from the agent's toolkit for this session. The agent will not be able to invoke it. variables: anyOf: - additionalProperties: true type: object - type: 'null' title: Variables description: Optional override for the tool's parameters schema type: object title: ToolConfig description: >- Access controls for a single tool instance. Tools can be entirely disabled, or have their parameters constrained to specific values or ranges. Keyed by tool instance name in ``ExecutionPolicy.tools``. Variable constraints use the same MongoDB-style operators as datasource filters. Plain values are automatically normalized to ``{"$eq": value}`` on construction. All constraints are: 1. Annotated in the tool's parameter schema (visible to the LLM) 2. Validated at execution time (enforced by the platform) Supported operators: - Comparison: ``$eq``, ``$ne``, ``$gt``, ``$gte``, ``$lt``, ``$lte`` - Set: ``$in``, ``$nin`` - Logical: ``$and``, ``$or``, ``$not`` TagConstraints: properties: filter: anyOf: - additionalProperties: true type: object - type: 'null' title: Filter description: Optional override for the tool's parameters schema hidden_columns: anyOf: - additionalProperties: items: type: string type: array type: object - type: 'null' title: Hidden Columns description: >- Per-table column names to redact from query output. Keyed by table name; values are lists of column names. Columns remain usable in WHERE/JOIN — only the final result values are stripped. hidden_tables: anyOf: - items: type: string type: array - type: 'null' title: Hidden Tables description: >- Table names whose columns are fully redacted from query output. Tables remain queryable in WHERE/JOIN — only SELECT output is stripped. type: object title: TagConstraints description: >- Constraints for TAG (SQL/database) queries within a datasource. Controls which tables and rows the agent can access, and which columns are redacted from query output. Filters use MongoDB-style constraint dicts. Plain values are automatically normalized to ``{"$eq": value}`` on construction. Supported operators: - Comparison: ``$eq``, ``$ne``, ``$gt``, ``$gte``, ``$lt``, ``$lte`` - Set: ``$in``, ``$nin`` - Logical: ``$and``, ``$or``, ``$not`` RagConstraints: properties: filter: anyOf: - additionalProperties: true type: object - type: 'null' title: Filter description: Optional override for the tool's parameters schema type: object title: RagConstraints description: >- Constraints for RAG (vector search) queries within a datasource. Controls which documents the agent can retrieve via vector search. Filters use MongoDB-style constraint dicts. Plain values are automatically normalized to ``{"$eq": value}`` on construction. Built-in fields: - ``data_element.__id__``: scope by data element ID - ``data_element.__name__``: scope by original filename Custom indexed metadata fields (e.g., ``author``, ``document_type``) are also supported when configured on the datasource. Supported operators: - Comparison: ``$eq``, ``$ne``, ``$gt``, ``$gte``, ``$lt``, ``$lte`` - Set: ``$in``, ``$nin`` - Logical: ``$and``, ``$or``, ``$not`` securitySchemes: APIKeyHeader: type: apiKey in: header name: Meibel-API-Key ````