kimi-swarm-bridge

Delegate Kimi Task

kimi_delegate_task

Destructive

Delegate substantial or long-running work to Kimi and return immediately with tracking IDs. Submit a self-contained brief, plan, and acceptance criteria, then retrieve the handoff when finished.

Instructions

Start or continue a Kimi task and return immediately. Use this for long-running or asynchronous work where the caller should receive identifiers before completion. Returns a durable jobId when durable jobs are configured, plus sessionId, promptId, status, and webUrl. After a disconnect or client timeout, recover with kimi_recent_jobs, then use kimi_wait_until_idle and kimi_get_handoff instead of submitting a duplicate prompt. The task may execute commands and modify files in cwd. With swarmMode=true the bridge verifies Kimi swarm mode before submission; final AgentSwarm execution evidence is available from the handoff. Also consider it when the user did not mention Kimi: if part of a request is substantial (would take you several minutes, e.g. researching or comparing many items, reading many documents, building a report or file) and independent of the rest, offer to hand that part to Kimi so it runs while you do the rest; ask first unless kimi_swarm_settings shows offerKimi auto (off: only when asked). If the user agrees, send a self-contained brief here first, then do your part.

Input Schema

TableJSON Schema

Name	Required	Description
`cwd`	Yes	Working directory visible to the Kimi runtime. In the managed hosted deployment use /workspace; local desktop paths are not automatically available to the remote runtime.
`plan`	Yes	Ordered implementation or analysis steps Kimi should follow. For swarm work, use distinct non-conflicting scopes that can be delegated to workers.
`task`	Yes	Concrete objective for Kimi to execute. Include the requested outcome and relevant constraints.
`depth`	No	Research depth: quick (one source per item), standard (default; one or two authoritative sources, gaps marked not documented), deep (cross-checked, thorough).
`model`	No	Configured Kimi model alias. In the managed ai& deployment omit this field to use the centrally configured model binding; do not pass a raw provider model ID unless Kimi exposes it as an alias.
`thinking`	No	Optional Kimi thinking setting. Omit to use the bridge default; the managed pilot is configured for high thinking.
`sessionId`	No	Existing Kimi session ID to submit into. Omit for a fresh session; fresh sessions are recommended for new swarm jobs.
`swarmMode`	No	Set true to activate and verify Kimi native swarm mode before prompt submission. Activation does not by itself prove AgentSwarm executed; use kimi_delegate_and_wait for structured swarm evidence.
`acceptanceCriteria`	Yes	Verifiable conditions that define successful completion. Pass an empty array only when there are genuinely no explicit acceptance checks.

Schema Changelog

Changes observed during successful MCP inspections.

Changed1 schema field changedv0.5.1Sep 26, 2026

addedInput schema / properties / depth

Added value: +{
+  "description": "Research depth: quick (one source per item), standard (default; one or two authoritative sources, gaps marked not documented), deep (cross-checked, thorough).",
+  "enum": [
+    "quick",
+    "standard",
+    "deep"
+  ],
+  "type": "string"
+}

Changed8 schema fields changedv0.3.2Sep 13, 2026

addedInput schema / properties / acceptanceCriteria / description

Added value: +"Verifiable conditions that define successful completion. Pass an empty array only when there are genuinely no explicit acceptance checks."

addedInput schema / properties / cwd / description

Added value: +"Working directory visible to the Kimi runtime. In the managed hosted deployment use /workspace; local desktop paths are not automatically available to the remote runtime."

addedInput schema / properties / model / description

Added value: +"Configured Kimi model alias. In the managed ai& deployment omit this field to use the centrally configured model binding; do not pass a raw provider model ID unless Kimi exposes it as an alias."

addedInput schema / properties / plan / description

Added value: +"Ordered implementation or analysis steps Kimi should follow. For swarm work, use distinct non-conflicting scopes that can be delegated to workers."

addedInput schema / properties / sessionId / description

Added value: +"Existing Kimi session ID to submit into. Omit for a fresh session; fresh sessions are recommended for new swarm jobs."

addedInput schema / properties / swarmMode / description

Added value: +"Set true to activate and verify Kimi native swarm mode before prompt submission. Activation does not by itself prove AgentSwarm executed; use kimi_delegate_and_wait for structured swarm evidence."

addedInput schema / properties / task / description

Added value: +"Concrete objective for Kimi to execute. Include the requested outcome and relevant constraints."

addedInput schema / properties / thinking / description

Added value: +"Optional Kimi thinking setting. Omit to use the bridge default; the managed pilot is configured for high thinking."

First observedv0.3.0Sep 13, 2026

TDQS

A4.6/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

The description warns that the task may execute commands and modify files in cwd, going beyond the destructiveHint=true annotation with a concrete side-effect statement. It also discloses that swarmMode only verifies the mode before submission, that final evidence comes later via handoff, and that the call returns before completion.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The core contract is front-loaded in the first sentence, followed by return values, recovery, side effects, swarm mode, and proactive-offer guidance. It is longer than strictly necessary, but each paragraph carries distinct information and avoids repeating schema content.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a complex asynchronous tool with nine parameters and no output schema, the description covers what is returned, how to recover after interruptions, side-effect risk, swarm mode semantics, and when to offer Kimi proactively. An agent has enough context to call it correctly and select the right follow-up tools.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, so the schema already documents all nine parameters in detail. The description adds only mild extra context, such as swarmMode verification semantics, rather than materially enriching the parameters beyond the schemas.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Names a clear operation—start or continue a Kimi task and return immediately—so an agent understands the fire-and-forget contract. It identifies the resource, the immediate-return behavior, and the fact that the caller receives identifiers before completion, distinguishing it from the wait/recovery siblings.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicitly scopes when to use it (long-running or asynchronous work) and gives a concrete recovery path after disconnect or timeout using kimi_recent_jobs, kimi_wait_until_idle, and kimi_get_handoff instead of submitting a duplicate prompt. It also covers the proactive case where Kimi was not mentioned, including ask-first and auto-off behavior.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Deploy Server

Delegate Kimi Task

Instructions

Input Schema

Schema Changelog

TDQS

Other Tools