Multi-agent delegation

The middleware package provides agents() for delegation. It injects one delegation tool per sub-agent. By default, tool names use delegate_to_<agentName>.

import { agents, artifacts, retry } from '@genkit-ai/middleware';

const researcher = ai.defineAgent({
  name: 'researcher',
  description: 'Finds facts and produces sourced research notes.',
  system: 'Research the user request and write concise findings.',
  use: [artifacts(), retry()],
});

const coder = ai.defineAgent({
  name: 'coder',
  description: 'Writes and explains code.',
  system:
    'Write clear TypeScript code unless the user asks for another language.',
  use: [artifacts(), retry()],
});

const coordinator = ai.defineAgent({
  name: 'coordinator',
  system:
    'Delegate to specialists, inspect their results, then answer the user.',
  use: [
    agents({
      agents: [
        'researcher',
        {
          name: 'coder',
          description:
            'Writes, debugs, and explains code. Use for programming tasks.',
        },
      ],
      historyLength: 4,
      maxDelegations: 5,
      artifactStrategy: 'session',
    }),
    artifacts({ readonly: true }),
  ],
});

The middleware can discover agent descriptions from action metadata, or you can override a description in the middleware config. Keep descriptions concrete because they become tool descriptions for the orchestrator model.

• agents accepts agent names, agent actions, or entries with a name and description override.
• toolPrefix controls generated tool names. It defaults to delegate_to; set it to an empty string to use bare agent names.
• historyLength sets how many recent conversation messages are forwarded to sub-agents.
• maxDelegations limits delegation calls in one orchestrator turn.
• artifactStrategy controls whether sub-agent artifacts are merged into the parent session.

When history is forwarded to client-managed sub-agents, the middleware includes recent messages in the sub-agent state. For server-managed sub-agents, history is not forwarded as client state because those agents own their server-side session.

Delegation appears as normal tool activity in the orchestrator stream.

const turn = coordinator
  .chat()
  .sendStream('Research sorting algorithms and write quicksort.');

for await (const chunk of turn.stream) {
  for (const request of chunk.toolRequests) {
    const name = request.toolRequest.name;
    if (name.startsWith('delegate_to_')) {
      showDelegation(name);
    }
  }

  if (chunk.text) {
    appendText(chunk.text);
  }
}

Sub-agent interrupts and failures are returned to the orchestrator as tool output. They do not automatically become top-level interrupts for the original client. Write orchestrator instructions that tell it how to handle delegated failures, such as retrying, choosing another specialist, or asking the user for clarification.

With artifactStrategy: 'session', sub-agent artifacts are merged into the parent session and namespaced by invocation. Pair this with artifacts({ readonly: true }) so the orchestrator can inspect delegated work through the read_artifact tool.

Use session artifacts when delegated work should be visible to the final user or to later turns. Keep artifacts isolated when the specialist output is only an implementation detail for the orchestrator’s current answer.

The older Building multi-agent systems page describes a prompts-as-tools pattern. Prefer the Agents API middleware for new work because it integrates with sessions, streaming, persistence, background execution, and HTTP clients.

The experimental middleware package github.com/firebase/genkit/go/plugins/middleware/exp provides Agents for delegation. It injects one delegation tool per sub-agent (named delegate_to_<agentName> by default) and appends a <sub-agents> listing to the orchestrator’s system prompt. Attach it with ai.WithUse inside the agent’s inline prompt.

import (
    "github.com/firebase/genkit/go/ai"
    aix "github.com/firebase/genkit/go/ai/exp"
    "github.com/firebase/genkit/go/ai/exp/localstore"
    "github.com/firebase/genkit/go/genkit"
    genkitx "github.com/firebase/genkit/go/genkit/exp"
    middlewarex "github.com/firebase/genkit/go/plugins/middleware/exp"
    "github.com/firebase/genkit/go/plugins/googlegenai"
)

The snippets below share one Genkit instance and one store for the orchestrator:

g := genkit.Init(ctx,
    genkit.WithExperimental(), // Required: the exp constructors panic without it.
    genkit.WithPlugins(&googlegenai.GoogleAI{}),
)

store := localstore.NewInMemorySessionStore[any]()

researcher := genkitx.DefineAgent(g, "researcher",
    aix.InlinePrompt{
        ai.WithModelName("googleai/gemini-flash-latest"),
        ai.WithSystem("Research the user request and write concise findings."),
        ai.WithUse(&middlewarex.Artifacts{}),
    },
    aix.WithDescription[any]("Finds facts and produces sourced research notes."),
)

coder := genkitx.DefineAgent(g, "coder",
    aix.InlinePrompt{
        ai.WithModelName("googleai/gemini-flash-latest"),
        ai.WithSystem("Write clear Go code unless the user asks for another language."),
        ai.WithUse(&middlewarex.Artifacts{}),
    },
    aix.WithDescription[any]("Writes, debugs, and explains code. Use for programming tasks."),
)

coordinator := genkitx.DefineAgent(g, "coordinator",
    aix.InlinePrompt{
        ai.WithModelName("googleai/gemini-flash-latest"),
        ai.WithSystem("Delegate to specialists, inspect their results, then answer the user."),
        ai.WithUse(
            &middlewarex.Agents{
                Agents:           []aix.AgentRef{researcher.Ref(), coder.Ref()},
                HistoryLength:    4,
                MaxDelegations:   5,
                ArtifactStrategy: middlewarex.ArtifactStrategySession,
            },
            &middlewarex.Artifacts{Readonly: true},
        ),
    },
    aix.WithSessionStore(store),
)

Reference a sub-agent by name (aix.AgentRef{Name: "researcher"}) or capture it from an agent value with agent.Ref(), which carries the agent’s description into the system listing. Descriptions matter because they become the delegation tool descriptions the orchestrator model sees, so keep them concrete.

The middleware resolves sub-agents through the Genkit instance seeded on the turn context, which genkitx.DefineAgent (and genkit.Generate) set automatically. Attach it to the orchestrator agent. Delegation composes: a sub-agent that carries its own Agents middleware delegates further, so orchestrations nest without extra wiring.

Middleware is per-agent. A delegation tool runs the sub-agent as its own invocation, so middleware attached with ai.WithUse on the orchestrator’s inline prompt wraps only the orchestrator’s model calls, never a sub-agent’s. Attach cross-cutting middleware such as redaction or logging to every agent that should have it. Doing so does not double-apply: the two agents’ generate calls are separate.

The orchestrator agent in the basic-agents sample is this arrangement running: it delegates to two client-managed sub-agents and reads their work back through session artifacts.

The Agents middleware is configured through struct fields:

• Agents lists the sub-agents available for delegation, by name or via agent.Ref(). At least one is required.
• ToolPrefix controls generated tool names. A nil value defaults to delegate_to (tools become delegate_to_<agent>); a pointer to the empty string uses bare agent names. A non-empty prefix also namespaces the shared tools described below, so two Agents instances in one generate call need distinct, explicit prefixes.
• MaxDelegations caps delegation calls in one orchestrator generate call. 0 means unlimited. Background launches and continuations spend the same budget.
• HistoryLength sets how many recent conversation messages are forwarded to a sub-agent. 0 forwards only the task description.
• ArtifactStrategy controls how sub-agent artifacts surface, ArtifactStrategyInline (default) or ArtifactStrategySession.
• Async lets the orchestrator launch a sub-agent in the background and collect its result later. See Delegate in the background.

History is forwarded only to client-managed sub-agents (those without a session store). A server-managed sub-agent owns its server-side session, so it receives only the task description.

Every delegation tool takes a task and an optional name, a short label the middleware echoes on the result and on background-task reports next to the task ID. It is a reading aid for a model juggling several delegations, not an identifier.

An orchestrator that delegates and waits is blocked for as long as the sub-agent runs. Set Async: true to let it keep working instead. Every delegation tool then takes a background flag that starts the sub-agent through its detach support and returns a task ID at once, and three shared tools give the orchestrator one control per thing it can do with a launched task.

It is the sub-agent that needs a session store here, one whose store supports detach: background work is tracked by a pending snapshot, so a sub-agent without one can only be delegated to synchronously, and the launch is refused with a hint to retry that way. The orchestrator itself needs no store for this.

logAnalyst := genkitx.DefineAgent(g, "log_analyst",
    aix.InlinePrompt{
        ai.WithModelName("googleai/gemini-flash-latest"),
        ai.WithSystem("Scan the logs of the service you are given and report the failure signature."),
        ai.WithTools(queryLogs),
    },
    aix.WithSessionStore(localstore.NewInMemorySessionStore[any]()),
    aix.WithDescription[any]("Scans service logs and reports the failure signature. Slow: a scan takes tens of seconds."),
)

commander := genkitx.DefineAgent(g, "commander",
    aix.InlinePrompt{
        ai.WithModelName("googleai/gemini-flash-latest"),
        ai.WithSystem("Run the incident. Start every investigation in the background, post a status update at once, then wait for the results."),
        ai.WithTools(postStatus),
        // Launch, post, wait, and post again are all tool rounds.
        ai.WithMaxTurns(15),
        ai.WithUse(&middlewarex.Agents{
            Agents:         []aix.AgentRef{logAnalyst.Ref()},
            Async:          true,
            MaxDelegations: 6,
        }),
    },
    aix.WithSessionStore(store),
)

The orchestrator launches with {"task": "...", "background": true}, calls other tools while the sub-agent runs, and collects the result later. wait_for_background_tasks takes an optional timeoutSeconds, so a slow task becomes an interim answer instead of a blocked turn: zero waits until the tasks settle, and a wait that runs out returns the current statuses with timedOut set. waitFor: "first" turns the join into a race that returns as soon as any listed task settles while the rest keep running. An abort is safe on any task and never blocks: a task that had already finished is left alone and reports its result, and a live one reports aborting while it saves its progress, settling as a resumable aborted that the wait tool can collect.

The middleware keeps no task registry. A task ID is <agent>:<snapshotId>, and it rides in the tool result, so the orchestrator’s own conversation is the registry: a re-instantiated orchestrator collects with nothing but the IDs in its history. Reports key on the snapshot. A pending task reports its status alone; a completed one carries the sub-agent’s last message and its artifacts; a failed, aborted, or expired one carries an explanatory error; a completed run whose finish reason carries no answer, such as an interrupt, reports as failed with the reason. A task ID the store cannot find reports “delegate again”.

Two prompting details matter. The middleware explains how background delegation works, but not when to use it, so tell the orchestrator which delegations to background and when to post interim updates. And raise ai.WithMaxTurns: launching, posting, waiting, and posting again are all tool rounds, so an orchestrator that collects in the background needs more room than one that blocks on each delegation. The commander agent in the basic-agents sample is the worked example: an incident commander that starts two slow investigators in the background, posts its first update while they run, and folds their answers in as they settle.

Every delegation to a server-managed sub-agent leaves a continuable handle behind. A synchronous result names the run’s last committed snapshot as taskId with its settled status, and background reports for failed, aborted, and expired tasks name the same handle. The shared continue_task tool spends it:

• A failed or aborted task continues from its last saved progress. With no instructions the committed turn is re-attempted as it stood; with instructions the retry is steered by a fresh user message.
• A completed task accepts follow-up instructions inside the sub-agent’s own session, so pressing on never repeats finished work. It is refused without them, since an empty input would re-run the finished turn.
• An expired task, whose worker died, is fenced with an abort and continued from its parent snapshot, the last one committed before the launch. A launch that never committed a turn has nothing saved, and the tool says to delegate again.
• A task that stopped on an interrupt is refused as a dead end, since continuing it would mean answering the interrupt. The orchestrator delegates a more self-contained task instead.

With Async set, continue_task also takes background: true and returns a fresh task ID in the same session. A client-managed delegation settles inline, carries no taskId, and is redone by delegating again. The tool registers only when some configured sub-agent may be server-managed, so an all-client-managed configuration gets no dead tool. A continuation spends a MaxDelegations slot; a refusal that names a retry which can succeed refunds it.

The Artifacts middleware gives a model read_artifact and write_artifact tools over the active session’s artifacts, and injects an <artifacts> listing into the system prompt each turn. Set Readonly: true to provide only read_artifact.

With ArtifactStrategySession, a sub-agent’s artifacts are merged into the parent session and kept out of the tool result. They are namespaced by invocation, <agent>_<snapshotId prefix>/<name> for a run with a snapshot behind it and <agent>_<n>/<name> otherwise, so a later check of the same task overwrites its earlier merge rather than duplicating it. Pair the strategy with &middlewarex.Artifacts{Readonly: true} on the orchestrator so it can inspect delegated work through read_artifact before answering. The default ArtifactStrategyInline instead includes artifact content in the delegation tool result and also merges it into the session.

Artifacts live on the active agent session, so the Artifacts tools only have an effect inside an agent invocation. With no active session they degrade gracefully: the listing is empty and the tools report that.

A sub-agent failure is returned to the orchestrator as the delegation tool’s output, with the task ID to continue it, rather than propagated as a top-level error to the original client. A sub-agent interrupt is reported the same way but cannot be continued: there is no stateful sub-agent runtime to answer it from. Write orchestrator instructions that say how to handle a delegated failure, such as continuing the task, choosing another specialist, or asking the user for clarification.

Task handles are not access-scoped. The background-task and continue tools read any snapshot ID belonging to a configured sub-agent, whether or not this conversation launched it, mirroring the sub-agent’s own companion actions. In a multi-tenant deployment treat snapshot IDs as capability-like secrets: text that reaches the orchestrator model can steer these tools at any ID it names.

Using the middlewares through ai.WithUse needs no plugin. Register &middlewarex.Middleware{} only to make them resolvable by name, for example in the Developer UI.

g := genkit.Init(ctx,
    genkit.WithExperimental(),
    genkit.WithPlugins(&googlegenai.GoogleAI{}, &middlewarex.Middleware{}),
)

• Sessions and state covers the session artifacts a sub-agent writes into.
• Background execution covers the detached runs that back background delegation.
• Agent error handling covers what a delegated failure looks like to the orchestrator.
• Custom orchestration covers taking over the turn loop when middleware delegation is not enough.

Genkit Dart provides the agents() middleware from package:genkit_middleware/agents.dart to manage multi-agent delegation. It dynamically auto-injects one delegation tool per sub-agent (named delegate_to_<agentName> by default) and appends a list of available sub-agents and their descriptions to the orchestrator’s system prompt.

Make sure to include AgentsPlugin() in your Genkit initialization.

The agents() middleware is experimental and pulls getCurrentSession from package:genkit/experimental.dart, which is also where defineAgent() lives.

import 'package:genkit/genkit.dart';
import 'package:genkit/experimental.dart';
import 'package:genkit_middleware/agents.dart';

final researcher = ai.defineAgent(
  name: 'researcher',
  description: 'A thorough research assistant that provides well-sourced answers.',
  system: 'You are a thorough research assistant. Return a clear and factual answer.',
  maxTurns: 10,
);

final coder = ai.defineAgent(
  name: 'coder',
  description: 'Writes, debugs, and explains code.',
  system: 'You are an expert programmer. Use Dart by default.',
  maxTurns: 10,
);

final orchestratorAgent = ai.defineAgent(
  name: 'orchestratorAgent',
  system: '''
    You are a project assistant. Analyze the user's request and delegate to the appropriate sub-agent.
    If the request requires both research and code, call them sequentially.
    After receiving sub-agent responses, synthesize a final answer for the user.
  ''',
  use: [
    agents(
      agents: ['researcher', 'coder'],
      maxDelegations: 5,
      historyLength: 4,
    ),
  ],
  store: InMemorySessionStore(),
);

Always provide a clear, descriptive description for sub-agents, as this metadata is used directly by the orchestrator model to determine when to call each delegation tool.

• agents is a list of sub-agent names available to the orchestrator.
• toolPrefix controls generated tool names. It defaults to delegate_to (tools become delegate_to_<agent>); set it to an empty string to use bare agent names.
• maxDelegations caps delegation calls in one orchestrator turn to prevent runaway loops (e.g., 5).
• historyLength sets how many recent conversation messages are forwarded to the sub-agents so they have context. 0 or omitted forwards only the task description.
• artifactStrategy controls how sub-agent artifacts surface: inline (default) includes artifact content in the delegation tool result and merges artifacts into the parent session; session merges into the parent session only, so the tool result names the artifacts without their content.
• async enables background delegation. See Delegate in the background.

History is forwarded only to client-managed sub-agents (those without a session store). A server-managed sub-agent owns its server-side session, so it receives only the task description.

Delegation appears as a standard tool call in the orchestrator’s chunk stream. This allows clients to see in real-time which sub-agent is active.

final turn = orchestratorAgent.chat().sendStream(text: 'Research quicksort and write it in Dart.');

await for (final chunk in turn.stream) {
  for (final req in chunk.toolRequests) {
    final name = req.toolRequest.name;
    if (name.startsWith('delegate_to_')) {
      print('Delegating to sub-agent: $name');
    }
  }
  if (chunk.text.isNotEmpty) {
    stdout.write(chunk.text);
  }
}

An orchestrator that delegates and waits is blocked for as long as the sub-agent runs. Set async: true to let it keep working instead. Each delegation tool then accepts a background flag that starts the sub-agent and returns a task ID at once, and three shared tools give the orchestrator one control per thing it can do with a launched task.

Background delegation requires server-managed sub-agents: the launched sub-agent needs a session store that supports detach, because background work is tracked by a pending snapshot. A sub-agent without one can only be delegated to synchronously, and the launch is refused with a hint to retry that way. The orchestrator itself needs no store for this.

import 'package:genkit/genkit.dart';
import 'package:genkit/experimental.dart';
import 'package:genkit_middleware/agents.dart';

// Server-managed: the store is what makes background delegation possible.
final researcher = ai.defineAgent(
  name: 'researcher',
  description: 'Researches a single, focused topic and returns a concise summary.',
  system: 'You are a focused research assistant. Given one topic, return a tight summary.',
  store: InMemorySessionStore(),
  maxTurns: 4,
);

final orchestrator = ai.defineAgent(
  name: 'asyncOrchestrator',
  system: '''
    When the user asks about several topics, delegate each one to "researcher"
    in the background (set "background": true) so they run in parallel, keeping
    the taskId each delegation returns. Once every task is launched, call
    wait_for_background_tasks with all the taskIds to collect the results, then
    synthesize a final answer.
  ''',
  use: [
    agents(agents: ['researcher'], async: true, maxDelegations: 8),
  ],
  store: InMemorySessionStore(),
);

The orchestrator launches with {"task": "...", "background": true}, calls other tools while the sub-agent runs, and collects the result later. wait_for_background_tasks takes an optional timeoutSeconds, so a slow task becomes an interim answer instead of a blocked turn: on timeout the current statuses are returned. An abort is safe on any task and never blocks: a task that already finished is left alone and reports its result, and a live one reports aborting while it saves progress, settling as a resumable aborted.

Task IDs ride in the tool result, so the orchestrator’s own conversation is the registry: task IDs from earlier tool results stay valid across turns, and re-delegating the same work is avoidable by checking them first. Because launching, collecting, and following up are separate tool rounds, raise maxTurns and maxDelegations so a background-collecting orchestrator has room to work.

The middleware explains how background delegation works, but not when to use it, so tell the orchestrator which delegations to background and when to post interim updates. The async agent sample is the worked example: an orchestrator that fans several research tasks out in the background and folds their answers in as they settle.

Every delegation to a server-managed sub-agent leaves a continuable handle behind, named by the task’s taskId. The continue_task tool spends it:

• A failed or aborted task continues from its last saved progress. With no instructions the committed turn is re-attempted as it stood; with instructions the retry is steered by a fresh user message.
• A completed task accepts follow-up instructions inside the sub-agent’s own session, so pressing on never repeats finished work. It is refused without them.
• A task that stopped on an interrupt is refused as a dead end, since continuing it would mean answering the interrupt.

The tool registers only when a configured sub-agent may be server-managed, so an all-client-managed configuration gets no dead tool. With async: true, continue_task also accepts background: true and returns a fresh task ID in the same session.

If a sub-agent fails or triggers an interrupt, the failure or pause is returned to the orchestrator as the delegation tool’s output. It does not automatically bubble up as a top-level error to the client. You should instruct the orchestrator on how to handle failures, for example by continuing the task with continue_task, trying a different specialist, correcting input, or reporting the issue back to the user.