> ## Documentation Index > Fetch the complete documentation index at: https://docs.sidenet.ai/llms.txt > Use this file to discover all available pages before exploring further. # Save agent draft > Partial save of the draft agent config (instructions, model fields, agentTools, workflows, memory, scorers, subagents, etc). Only the fields you send are written. Subagents are `[{ agentId }]`. The model fields are `model_slug`, `policy_slug`, `model_options` and `model_settings`; the older `requesty_model_slug`, `requesty_policy_slug` and `requesty_options` spellings are still accepted and mapped onto them (the canonical name wins if you send both). `prompt`, `model_name`, `model_provider` and `model_openrouter_slug` no longer exist and are ignored — the system prompt is `instructions`. `name` and `description` are the agent's own, not the draft's: they are saved here for convenience but apply immediately to every version, published ones included, and need no publish. The response carries the updated `agent` alongside the `draft`. ## OpenAPI ````yaml /api-reference/openapi.json patch /v1/agents/{id}/draft openapi: 3.1.0 info: title: Sidenet API version: 1.0.0 description: >- Sidenet HTTP endpoints exposed by the Sidenet Studio. All routes require an api key that can be generated through the studio in studio.sidenet.ai. servers: - url: https://api.sidenet.ai security: - bearerAuth: [] paths: /v1/agents/{id}/draft: patch: tags: - Agents summary: Save agent draft description: >- Partial save of the draft agent config (instructions, model fields, agentTools, workflows, memory, scorers, subagents, etc). Only the fields you send are written. Subagents are `[{ agentId }]`. The model fields are `model_slug`, `policy_slug`, `model_options` and `model_settings`; the older `requesty_model_slug`, `requesty_policy_slug` and `requesty_options` spellings are still accepted and mapped onto them (the canonical name wins if you send both). `prompt`, `model_name`, `model_provider` and `model_openrouter_slug` no longer exist and are ignored — the system prompt is `instructions`. `name` and `description` are the agent's own, not the draft's: they are saved here for convenience but apply immediately to every version, published ones included, and need no publish. The response carries the updated `agent` alongside the `draft`. parameters: - name: id in: path required: true schema: type: string example: 6f1c0f4e-2b7a-4a51-9a2f-0c9d1b3e5a10 requestBody: required: true content: application/json: schema: type: object properties: instructions: type: array description: >- The agent's system prompt, as an ordered list of blocks. Each entry is one of: `{ type: "text", content }` — inline text; `{ type: "prompt_block", content, rules? }` — inline text with an optional display condition; `{ type: "prompt_block_ref", id }` — a reference to a /v1/prompt-blocks block in the same organization. Content may contain `{{variable}}`, `{{a.b}}` and `{{variable || 'default'}}` placeholders, resolved per request from the chat body's `variables` plus the `now.*` builtins; an unsupplied placeholder is left visible. `rules` is a `{ operator: "AND"|"OR", conditions: [...] }` group (max 3 levels) — the block is included only when it evaluates true. On the draft a ref floats to the block's active version; publishing PINS it, so a published agent version keeps its exact prompt until you republish (see /v1/prompt-blocks/:id/rollout). Pins are server-owned — do not send `versionId`. items: type: object example: - type: prompt_block_ref id: d41a8f6b-9c27-4b13-a5e8-2f60c1d7b394 - type: text content: Answer in {{locale || 'English'}}. model_slug: type: string description: >- The concrete model to run, e.g. `anthropic/claude-sonnet-4-5`. Ignored when `policy_slug` is set. example: anthropic/claude-sonnet-4-5 policy_slug: type: string description: >- A routing policy to run instead of a fixed model, e.g. `policy/fastest-sonnet`. Overrides `model_slug` when set. example: policy/fastest-sonnet model_options: type: object description: >- Model-level options passed through to the router — reasoning effort, native provider options, and the like. example: reasoningEffort: medium model_settings: type: object description: 'Generation settings: temperature, maxTokens, and similar.' example: temperature: 0.2 max_steps: type: integer minimum: 1 maximum: 100 description: >- Max LLM round-trips per run for the agentic tool loop. The final step is forced tool-free so runs always end with a text answer. null = platform default (20). example: 8 workflows: type: array description: >- Workflows the agent can run, as `{ workflow_id, config? }` entries. `workflow_id` is the `/v1/workflows` id and must belong to the same organization. The workflow's ACTIVE version is always what runs, so activating a new version reaches the agent on its next turn with no republish. The agent-facing tool name and its description come from the workflow itself — the model calls it as `workflow-`. The whole workflow runs inside one tool call, and its step progress is streamed to the client as `data-tool-workflow` / `data-tool-workflow-step` parts by /v1/chat. items: type: object required: - workflow_id properties: workflow_id: type: string config: type: object example: - workflow_id: 4c2f9e18-7a63-4d05-b1e8-93a7c0f2d641 agentTools: type: array description: >- Tools assigned to the agent. Each entry is `{ tool_id, provider_id?, config?, composio_connected_account_id?, discovery?, loop_responses?, require_approval? }`. `discovery` (integration tools only) controls where the tool schema lives at runtime: `true` puts it behind the agent's search_tools/load_tool discovery tools, `false` pins it into the prompt, and unset lets the platform decide — small toolkits stay in the prompt, large ones (beyond ~15 total tools) become discoverable so the prompt is not flooded with schemas. `require_approval: true` pauses every call to that tool for human approval: the chat stream emits a `data-tool-call-approval` chunk instead of executing (carrying a model-written `reason` sentence for the approval card), and the client continues the run by POSTing /v1/chat with the `approval` field. Set `loop_responses: true` on a tool to let one call cover a whole result set: it gains `_fields` (project only the fields needed), `_all_pages` (fetch and merge every page) and `_for_each` (call once per value of a parameter and merge the rows), and the agent is given the large-response retrieval tools automatically. Because those re-issue the request, enable it only on tools whose endpoint is safe to call repeatedly. items: type: object example: - tool_id: c07f2b95-8d41-4e63-a029-5f7b1c8d3e40 provider_id: 8d3b1a75-6c02-4e59-b84f-27a9d5e10c63 name: type: string description: >- The agent's name. Agent-level, not versioned — it takes effect immediately, with no publish, and is what other agents see when this one is a subagent. example: Support agent description: type: string description: >- The agent's description. Agent-level, not versioned — same immediate effect as `name`. Send null to clear it. example: Answers billing and account questions scorers: type: array items: type: object example: [] agents: type: array description: Subagents this one may delegate to. items: type: object properties: agentId: type: string example: - agentId: 6f1c0f4e-2b7a-4a51-9a2f-0c9d1b3e5a10 memory: type: object example: lastMessages: 20 semanticRecall: false responses: '200': description: The saved draft, and the agent record it belongs to content: application/json: schema: type: object properties: agent: type: object description: >- The agent after the save — `name`/`description` reflect this call, and the version summary says whether the draft now differs from what is published. properties: id: type: string format: uuid org_id: type: string format: uuid name: type: string description: type: string nullable: true active_version_id: type: string format: uuid nullable: true description: Null until a published version is activated. created_at: type: string format: date-time updated_at: type: string format: date-time activeVersionNumber: type: integer nullable: true description: >- Version number currently serving traffic. Null until a published version is activated. latestPublishedVersionNumber: type: integer nullable: true description: >- Newest published version number. Higher than `activeVersionNumber` when a newer version is published but not activated. Null before the first publish. draftDirty: type: boolean description: >- True when the draft's config differs from the latest published version — i.e. there are unpublished edits. Fetch them with `?version=draft`. draft: type: object properties: id: type: string format: uuid agent_id: type: string format: uuid version_number: type: integer description: >- 0 is the mutable draft; every published version is 1 or above. rev: type: integer description: >- Server-owned change counter, bumped on every save of the draft. Read-only — useful to tell whether a draft moved since you last read it. instructions: type: array description: >- The system prompt, as ordered blocks. Entries are `{type:"text"}`, `{type:"prompt_block"}` or `{type:"prompt_block_ref"}`. On a published version each ref also carries the `versionId` it was frozen to at publish time; on the draft that pin is absent. items: type: object model_slug: type: string nullable: true description: >- The concrete model this version runs. Ignored when `policy_slug` is set. policy_slug: type: string nullable: true description: >- Routing policy to run instead of a fixed model. Overrides `model_slug`. model_options: type: object nullable: true description: >- Model-level options passed to the router (reasoning effort, native options). model_settings: type: object nullable: true description: 'Generation settings: temperature, maxTokens, etc.' max_steps: type: integer nullable: true agentTools: type: array description: Tool references this version can call. items: type: object workflows: type: array description: Attached workflows. items: type: object agents: type: array description: >- Subagents this version can delegate to, as `[{ agentId }]`. items: type: object scorers: type: array items: type: object memory: type: object nullable: true created_at: type: string format: date-time updated_at: type: string format: date-time description: >- The whole draft after the merge, not just the fields you sent. `rev` has been bumped by the DB. '400': description: Invalid body '404': description: Not found components: securitySchemes: bearerAuth: type: http scheme: bearer description: >- Organization API key, generated in studio.sidenet.ai. Backend only — never in a browser. ````