Specification Parameters
System & Integration
- Create Session API
- https://ark.cn-beijing.volces.com/api/v3/sessions · Must carry the API Key and authenticate via Bearer token
- Response mechanism
- Events API + SSE 流式 · Send messages to the Session via the Events API endpoint and receive response events in a streaming manner
Storage
- Artifact path
- /mnt/session/outputs/ · Directory of final artifacts written by the Agent; can be bound to your own TOS directory for long-term retention
- Environment constraint
- TOS Bucket 与服务同地域 · Your own TOS must be deployed in the same region as the Managed Agents service
Comparison with Similar Products
| Specifications | 本品 · Managed Agents | kimi-k3.1-pro | 火山方舟模型 API | qwen-audio-3.1-asr-flash | doubao-seed-2.1-turbo |
|---|---|---|---|---|---|
| Main controller | — | — | — | — | — |
| NPU | — | — | — | — | — |
| Memory / VRAM | — | — | — | — | — |
| Reference unit price | Request a Quote | Request a Quote | Request a Quote | Request a Quote | Request a Quote |
| Inventory | — | — | — | — | — |
Product introduction
Service positioning & workflow: Managed Agents provides a visual interface that supports creating, configuring, and testing Agents without writing API code; once polished, the configuration is integrated into business code in the form of an ID.
Core concepts: An Agent is a versioned, reusable agent configuration asset that defines the role, capabilities, behavioral boundaries, and operating rules; an Environment defines a runtime environment template with pre-installed dependencies, environment variables, and artifact storage; a Session is an independent Agent task run covering the complete lifecycle from initiation through execution to completion.
Configuration capabilities: When creating an Agent, you can choose a general Agent or a Seedance Managed Agent, bind a model, and write the system prompt; attach preset Skills from Volcano SkillHub or upload custom Skills locally; check built-in Agent Toolsets such as bash, edit, read, and write; connect remote MCP Servers via URL; and combine multiple Agents into collaborative workflows. Each configuration change automatically generates a new version, with version comparison, rollback, and gradual rollout supported.
Runtime environment: Environments currently support only the cloud-hosted type, with sandboxes spun up on demand by Ark and automatically put to sleep; pip and apt dependencies can be added as needed, and environment variables can be injected as key-value pairs; for long-term artifact retention, you must select an authorized TOS directory of your own, otherwise Ark's public TOS is used and files must be downloaded before the platform TTL expires; the TOS Bucket must be deployed in the same region as the service.
Debugging & integration: Each debugging session creates a new Session that displays an event timeline (Trace) covering the full pipeline — reasoning, tool calls, code execution, error retries, and more; upon completion, you can view final delivered artifacts, total duration, Token consumption, number of tool calls, and other statistics. To integrate into business code, simply copy the agent_id, environment_id, and API Key, send a request to the sessions endpoint to create a Session, then send messages via the Events API endpoint and receive response events as an SSE stream.
Pre-purchase checklist: This page does not disclose pricing, context length, or SLA; before integration, confirm in the console the activation requirements and billing terms for the models you need and for Seedance Managed Agent access.