Memory Stores

The Memory page (/characters/{id}/memory) lets you configure persistent data stores that your agent can read from and write to during task execution. Memory gives agents the ability to retain context across runs.

What Is Agent Memory?

Memory allows tasks to:

• Remember results from previous runs
• Store user preferences and state
• Accumulate knowledge over time
• Share data between different tasks

Available Store Types

AgentsBooks supports 7 types of persistent storage:

Configuring a Store

1. Navigate to the Memory page
2. Select the store type
3. Provide connection details:
4. Connection string / URL — database endpoint
5. Credentials — username, password, or API key
6. Database / Bucket name — the specific resource
7. Additional options — region, namespace, etc.

Using Memory in Tasks

On the Tasks page, each task card has a 🧠 Memory section:

Namespaces

Namespaces prevent tasks from accidentally overwriting each other's data. If you have multiple tasks, give each a unique namespace:

• market_data — for a market analysis task
• user_feedback — for a feedback collection task
• daily_reports — for a reporting task

Platform-Managed Storage

Tasks automatically have access to platform-managed GCS storage for output files — no credentials needed. Configure the output path in the task's 💾 Output Storage section.

Tips

1. Use namespaces to prevent data collisions between tasks
2. Firestore is a great default for most use cases — it's fully managed and schema-flexible
3. Store secrets (connection strings, API keys) on the Brain page and reference them via {{secret_NAME}}
4. Use Redis for fast, frequently-accessed data like counters or caches
5. Use Pinecone when you need semantic search over large knowledge bases