即梦 MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| capabilitiesA | 查询已实现操作、只读开关与积分上限;先查能力再执行。可调用不等于真实生成已验收。 |
| studio_catalogA | 读取网页角色库、剧本、故事板等功能的证据与缺口。此目录仅供规划,不能把 implemented=false 的条目当可调用工具。 |
| action_schemaA | 按 action 名查询一项固定操作的完整参数 schema。提交前查此工具,严格按真实字段填写。 |
| submit_jobA | 先用 action_schema 查询字段,再将 payload 提交为持久化 job。idempotencyKey 必填;超时复用原 key 和相同 payload。写入默认 dryRun=true,真实写入满足后端实时政策。COMPLETED 仅表示调用返回,须检查上游与素材状态;UNKNOWN 不得重新生成。 |
| submit_job_jsonA | 与 submit_job 完全等价的简洁输入入口。先查 action_schema,将含 action 的完整参数对象序列化为 payloadJson。后端仍逐字段严格校验;超时必须复用原 idempotencyKey,不自动重提。 |
| get_jobA | 只查询原jobId;status表示调用状态,operationOutcome表示业务结果。COMPLETED不代表媒体完成;partial/unknown按recovery核查原节点、资源或submitId,不重提。 |
| list_jobsA | 按状态、action或项目分页查询持久化任务摘要;不返回提示词或媒体内容。继续翻页须复用相同筛选和nextCursor。 |
| queue_summaryA | 查询持久化状态计数、读写队列和最早排队时间;不发起上游操作。 |
| cancel_jobA | 只原子取消QUEUED任务。已经开始的任务返回冲突,不发送任何上游取消或重试。重复取消原任务安全。 |
| artifact_manifestA | 核验已有受管成品并导出到本机私有目录,返回localPath、大小和SHA256。只接受项目与成品ID,不接受任意路径。无需HTTP或Bearer凭据。 |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 10 tools
Most tools are separable (list_jobs vs get_job vs queue_summary vs cancel_job), but submit_job and submit_job_json are explicitly described as completely equivalent entry points, which forces an arbitrary choice. The three discovery tools (capabilities, action_schema, studio_catalog) also overlap in intent and rely on prose to be told apart.
Consistent snake_case throughout, but conventions are mixed: verb_noun (list_jobs, submit_job, get_job, cancel_job) alongside bare noun phrases (capabilities, action_schema, studio_catalog, artifact_manifest, queue_summary). Still readable, but the pattern does not predict whether a tool is an action or a resource view.
Ten tools is a well-scoped size for a job submission/inspection server. The only redundancy is submit_job_json duplicating submit_job, which costs one slot but is documented as a convenience.
The lifecycle is largely covered: discovery (capabilities, action_schema, studio_catalog), submit, poll (get_job, list_jobs, queue_summary), cancel, and artifact export. Gaps are minor and arguably deliberate — no retry/resubmit tool and no job deletion or bulk cleanup.