Skip to main content
Memory lets your agent learn and improve across conversations. Deep Agents makes memory first class with filesystem-backed memory: the agent reads and writes memory as files, and you control where those files are stored using backends.
This page covers long-term memory: memory that persists across conversations. For short-term memory (conversation history and scratch files within a single session), see the context engineering guide. Short-term memory is managed automatically as part of the agent’s state.

​
How memory works

  1. Point the agent at memory files. Pass file paths to memory= when creating the agent. A backend controls where those files are stored and who can access them.
  2. Agent loads memory on startup. At the start of each conversation, the agent reads its memory files into the system prompt.
  3. Agent updates memory during the conversation. As the agent learns new information, it uses its built-in edit_file tool to update memory files. Changes are persisted and available in the next conversation.
The two most common patterns are agent-scoped memory (shared across all users) and user-scoped memory (isolated per user).

​
Agent-scoped memory

Give the agent a shared memory file that all users read from and write to. The agent improves over time as it accumulates knowledge across every conversation. It can also learn and update skills when it has write access. See skills as procedural memory. The key is the backend namespace: setting it to (assistant_id,) means every conversation for this agent reads and writes the same memory file. 
import { createDeepAgent, CompositeBackend, StateBackend, StoreBackend } from "deepagents";
import { getConfig } from "@langchain/langgraph";

const agent = createDeepAgent({
  memory: ["/memories/AGENTS.md"],
  backend: new CompositeBackend(
    new StateBackend(),
    {
      "/memories/": new StoreBackend({
        namespace: (ctx) => {
          const config = getConfig();
          return [config.metadata.assistantId];
        },
      }),
    },
  ),
});

​
User-scoped memory

Give each user their own memory file. The agent remembers preferences, context, and history per user while core agent instructions stay fixed. Users can also have per-user skills if stored in a user-scoped backend. The namespace uses (user_id,) so each user gets an isolated copy of the memory file. User A’s preferences never leak into User B’s conversations. 
import { createDeepAgent, CompositeBackend, StateBackend, StoreBackend } from "deepagents";

const agent = createDeepAgent({
  memory: ["/memories/preferences.md"],
  backend: new CompositeBackend(
    new StateBackend(),
    {
      "/memories/": new StoreBackend({
        namespace: (ctx) => [ctx.runtime.context.userId],
      }),
    },
  ),
});

​
Advanced usage

The sections above cover the basics: configure memory paths, choose a scope, and let the agent handle the rest. This section covers more advanced patterns.
DimensionQuestion it answersOptions
DurationHow long does it last?Short-term (single conversation) or long-term (across conversations)
Information typeWhat kind of information is it?Episodic (past experiences), procedural (instructions and skills), or semantic (facts)
ScopeWho can see and modify it?User, agent, or organization
Update strategyWhen are memories written?During conversation (default) or between conversations
RetrievalHow are memories read?Loaded into prompt (default) or on demand (e.g., skills)
Agent permissionsCan the agent write to memory?Read-write (default) or read-only (for shared policies)

​
Episodic memory

Episodic memory stores records of past experiences: what happened, in what order, and what the outcome was. Unlike semantic memory (facts and preferences stored in files like AGENTS.md), episodic memory preserves the full conversational context so the agent can recall how a problem was solved, not just what was learned from it. Deep Agents get episodic memory for free through checkpointers: every conversation is persisted as a checkpointed thread. To make past conversations searchable, wrap thread search in a tool. The user_id is pulled from the runtime context rather than passed as a parameter: 
import { Client } from "@langchain/langgraph-sdk";
import { getConfig } from "@langchain/langgraph";
import { tool } from "@langchain/core/tools";

const client = new Client({ apiUrl: "<DEPLOYMENT_URL>" });

const searchPastConversations = tool(
  async ({ query }: { query: string }) => {
    const config = getConfig();
    const userId = config.metadata.userId;
    const threads = await client.threads.search({
      metadata: { userId },
      limit: 5,
    });
    const results = [];
    for (const thread of threads) {
      const history = await client.threads.getHistory(thread.threadId);
      results.push(history);
    }
    return JSON.stringify(results);
  },
  {
    name: "search_past_conversations",
    description: "Search past conversations for relevant context.",
  }
);
You can scope thread search by user or organization by adjusting the metadata filter: 
// Search conversations for a specific user
const userThreads = await client.threads.search({
  metadata: { userId },
  limit: 5,
});

// Search conversations across an organization
const orgThreads = await client.threads.search({
  metadata: { orgId },
  limit: 5,
});
This is useful for agents that perform complex, multi-step tasks. For example, a coding agent can look back at a past debugging session and skip straight to the likely root cause.

​
Skills as procedural memory

Skills are a form of procedural memory: reusable instructions that tell the agent how to perform a task. Unlike semantic memory (facts) or episodic memory (experiences), procedural memory encodes step-by-step capabilities the agent can apply on demand. Skills can be either:
  • Read-only (developer-defined): The developer writes skills and the agent uses them but cannot modify them. This is the most common pattern.
  • Read-write (agent-learned): The agent creates and updates skills based on experience. When the agent has write access to memory, it can also write to its skills directory. Use policy hooks to control which paths are writable.
Skills are typically agent-scoped (shared across all users), but can also be user-scoped if stored in a user-namespaced backend. In Deep Agents, pass skills via the skills= parameter. Skills are loaded on demand rather than injected into every prompt, keeping context lean until a capability is needed: 
import { createDeepAgent } from "deepagents";

const agent = createDeepAgent({
  memory: ["/memories/AGENTS.md"],
  skills: ["/skills/"],
  // ...backend configuration
});
For a complete guide on defining, organizing, and using skills, see the Skills documentation.

​
Organization-level memory

Organization-level memory follows the same pattern as user-scoped memory, but with a shared namespace instead of a per-user one. Use it for policies or knowledge that should apply across all users and agents. Organization memory is typically read-only to prevent prompt injection via shared state. See read-only vs writable memory for details. 
import { createDeepAgent, CompositeBackend, StateBackend, StoreBackend } from "deepagents";

const agent = createDeepAgent({
  memory: [
    "/memories/preferences.md",
    "/policies/compliance.md",
  ],
  backend: new CompositeBackend(
    new StateBackend(),
    {
      "/memories/": new StoreBackend({
        namespace: (ctx) => [ctx.runtime.context.userId],
      }),
      "/policies/": new StoreBackend({
        namespace: (ctx) => [],
      }),
    },
  ),
});
Populate organization memory from your application code: 
import { Client } from "@langchain/langgraph-sdk";
import { createFileData } from "deepagents";

const client = new Client({ apiUrl: "<DEPLOYMENT_URL>" });

await client.store.putItem(
  [],
  "/compliance.md",
  createFileData(`## Compliance policies
- Never disclose internal pricing
- Always include disclaimers on financial advice
`),
);
Use policy hooks to enforce that org-level memory is read-only.

​
Background consolidation

By default, the agent writes memories during the conversation (hot path). An alternative is to process memories between conversations as a background task, sometimes called sleep time compute. A separate deep agent reviews recent conversations, extracts key facts, and merges them with existing memories.
ApproachProsCons
Hot path (during conversation)Memories available immediately, transparent to userAdds latency, agent must multitask
Background (between conversations)No user-facing latency, can synthesize across multiple conversationsMemories not available until next conversation, requires a second agent
For most applications, the hot path is sufficient. Add background consolidation when you need to reduce latency or improve memory quality across many conversations. All three approaches deploy a consolidation agent alongside your main agent: a deep agent that reads conversation history, extracts key facts, and merges them into the memory store.
TriggerWhen it runsBest for
CronFixed schedule (e.g., every 6 hours)Batching across many conversations, synthesizing trends
Scheduled runAfter a delay on the same threadSimple setups where the client controls when consolidation fires
All three approaches use the same consolidation agent and deployment configuration. Define these once, then choose a trigger.

​
Consolidation agent

The consolidation agent reads recent conversation history and merges key facts into the memory store. Register it alongside your main agent in langgraph.json:
src/consolidation-agent.ts
import { createDeepAgent } from "deepagents";
import { Client } from "@langchain/langgraph-sdk";
import { getConfig } from "@langchain/langgraph";
import { tool } from "@langchain/core/tools";

const sdkClient = new Client({ apiUrl: "<DEPLOYMENT_URL>" });

const searchRecentConversations = tool(
  async ({ query }: { query: string }) => {
    const config = getConfig();
    const userId = config.configurable.langgraph_auth_user_id;

    const since = new Date(Date.now() - 6 * 60 * 60 * 1000).toISOString();
    const threads = await sdkClient.threads.search({
      metadata: { userId },
      updatedAfter: since,
      limit: 20,
    });
    const conversations = [];
    for (const thread of threads) {
      const history = await sdkClient.threads.getHistory(thread.threadId);
      conversations.push(history.values.messages);
    }
    return JSON.stringify(conversations);
  },
  {
    name: "search_recent_conversations",
    description: "Search this user's conversations updated in the last 6 hours.",
  }
);

const agent = createDeepAgent({
  model: "claude-sonnet-4-6",
  systemPrompt: `Review recent conversations and update the user's memory file.
Merge new facts, remove outdated information, and keep it concise.`,
  tools: [searchRecentConversations],
});

export { agent };
langgraph.json
{
  "dependencies": ["."],
  "graphs": {
    "agent": "./src/agent.ts:agent",
    "consolidation_agent": "./src/consolidation-agent.ts:agent"
  },
  "env": ".env"
}

​
Cron

A cron job runs the consolidation agent on a fixed schedule. The agent searches all recent conversations and synthesizes them into memory. Schedule the consolidation agent with a cron job: 
import { Client } from "@langchain/langgraph-sdk";

const client = new Client({ apiUrl: "<DEPLOYMENT_URL>" });

const cronJob = await client.crons.create(
  "consolidation_agent",
  {
    schedule: "0 */6 * * *",
    input: { messages: [{ role: "user", content: "Consolidate recent memories." }] },
  },
);
All cron schedules are interpreted in UTC. See cron jobs for details on managing and deleting cron jobs.

​
Scheduled run

Schedule a run after each conversation using the after_seconds parameter. This is activity-driven: consolidation only fires when a user has been active. Pass the same thread_id so the consolidation agent has access to the conversation context. The delay gives the conversation time to finish (the user may send follow-up messages) so the consolidation agent sees the full exchange. If the user sends another message during the delay, the new run’s multitask_strategy controls what happens to the pending consolidation run. Use rollback on the new run to delete the stale consolidation run, then schedule a fresh one after the next response: 
import { Client } from "@langchain/langgraph-sdk";

const client = new Client({ apiUrl: "<DEPLOYMENT_URL>" });

// Schedule consolidation 30 minutes after the conversation
await client.runs.create(
  thread.thread_id,
  "consolidation_agent",
  {
    input: { messages: [{ role: "user", content: "Consolidate recent memories." }] },
    afterSeconds: 1800,
  },
);

// If the user sends another message, use rollback on the new run
// to delete the pending consolidation run before starting
const stream = client.runs.stream(
  thread.thread_id,
  "agent",
  {
    input: { messages: [{ role: "user", content: newMessage }] },
    multitaskStrategy: "rollback",
  },
);
See double texting for details on how multitask_strategy controls the behavior of concurrent runs on the same thread. For more on deploying agents with background processes, see going to production.

​
Read-only vs writable memory

By default, the agent can both read and write memory files. For shared state like organization policies or compliance rules, you may want to make memory read-only so the agent can reference it but not modify it. This prevents prompt injection via shared memory and ensures that only your application code controls what’s in the file.
PermissionUse caseHow it works
Read-write (default)User preferences, agent self-improvement, learned skillsAgent updates files via edit_file tool
Read-onlyOrganization policies, compliance rules, shared knowledge bases, developer-defined skillsPopulate via application code or the Store API. Use policy hooks to block agent writes.
Security considerations: If one user can write to memory that another user reads, a malicious user could inject instructions into shared state. To mitigate this:
  • Default to user scope (user_id) unless you have a specific reason to share
  • Use read-only memory for shared policies (populate via application code, not the agent)
  • Add human-in-the-loop validation before the agent writes to shared memory. Use an interrupt to require human approval for writes to sensitive paths.
To enforce read-only memory, use policy hooks on the backend to reject write operations to specific paths.

​
Concurrent writes

Multiple threads can write to memory in parallel, but concurrent writes to the same file can cause last-write-wins conflicts. For user-scoped memory this is rare since users typically have one active conversation at a time. For agent-scoped or organization-scoped memory, consider using background consolidation to serialize writes, or structure memory as separate files per topic to reduce contention. In practice, if a write fails due to a conflict, the LLM is usually smart enough to retry or recover gracefully, so a single lost write is not catastrophic.

​
Multiple agents in the same deployment

To give each agent its own memory in a shared deployment, add assistant_id to the namespace: 
new StoreBackend({
  namespace: (ctx) => {
    const config = getConfig();
    return [config.metadata.assistantId, ctx.runtime.context.userId];
  },
})
Use (assistant_id,) alone if you only need per-agent isolation without per-user scoping.
Use LangSmith tracing to audit what your agent writes to memory. Every file write appears as a tool call in the trace.
Edit this page on GitHub or file an issue.
Connect these docs to Claude, VSCode, and more via MCP for real-time answers.