Multi-Agent Shared Memory

Agents can share state via shared memory blocks. This allows agents to have a “shared memory”. You can shared blocks between agents by attaching the same block ID to multiple agents.

graph TD
    subgraph Supervisor
        S[Memory Block
I am a supervisor]
        SS[Shared Memory Block
Organization: Letta]
    end

    subgraph Worker
        W1[Memory Block
I am a worker]
        W1S[Shared Memory Block
Organization: Letta]
    end

    SS -..- W1S

In the example code below, we create a shared memory block and attach it to a supervisor agent and a worker agent. Because the memory block is shared, when one agent writes to it, the other agent can read the updates immediately.

TypeScript

// install letta-client with `npm install @letta-ai/letta-client`
import { LettaClient } from "@letta-ai/letta-client";

// create a client to connect to Letta
const client = new LettaClient({
  token: "LETTA_API_KEY",
});

// create a shared memory block
const sharedBlock = await client.blocks.create({
  label: "organization",
  description: "Shared information between all agents within the organization.",
  value: "Nothing here yet, we should update this over time.",
});

// create a supervisor agent
const supervisorAgent = await client.agents.create({
  model: "anthropic/claude-3-5-sonnet-20241022",
  embedding: "openai/text-embedding-3-small",
  // blocks created for this agent
  memoryBlocks: [{ label: "persona", value: "I am a supervisor" }],
  // pre-existing shared block that is "attached" to this agent
  blockIds: [sharedBlock.id],
});

// create a worker agent
const workerAgent = await client.agents.create({
  model: "anthropic/claude-3-5-sonnet-20241022",
  embedding: "openai/text-embedding-3-small",
  // blocks created for this agent
  memoryBlocks: [{ label: "persona", value: "I am a worker" }],
  // pre-existing shared block that is "attached" to this agent
  blockIds: [sharedBlock.id],
});

Python

python

# install letta_client with `pip install letta-client`
from letta_client import Letta

# create a client to connect to Letta
client = Letta(token="LETTA_API_KEY")

# create a shared memory block
shared_block = client.blocks.create(
    label="organization",
    description="Shared information between all agents within the organization.",
    value="Nothing here yet, we should update this over time."
)

# create a supervisor agent
supervisor_agent = client.agents.create(
    model="anthropic/claude-3-5-sonnet-20241022",
    embedding="openai/text-embedding-3-small",
    # blocks created for this agent
    memory_blocks=[{"label": "persona", "value": "I am a supervisor"}],
    # pre-existing shared block that is "attached" to this agent
    block_ids=[shared_block.id],
)

# create a worker agent
worker_agent = client.agents.create(
    model="anthropic/claude-3-5-sonnet-20241022",
    embedding="openai/text-embedding-3-small",
    # blocks created for this agent
    memory_blocks=[{"label": "persona", "value": "I am a worker"}],
    # pre-existing shared block that is "attached" to this agent
    block_ids=[shared_block.id],
)

Memory blocks can also be accessed by other agents, even if not shared. For example, worker agents can write the output of their task to a memory block, which is then read by a supervisor agent. To access the memory blocks of other agents, you can simply use the SDK clients or API to access specific agent’s memory blocks (using the core memory routes).