Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.litellm-agent-platform.ai/llms.txt

Use this file to discover all available pages before exploring further.

Memories are per-agent text snippets injected into the agent’s context at session start. They are intended for facts, preferences, and instructions that the agent should remember across sessions — for example, a user’s coding style or a recurring project convention. Memories can be searched with a keyword query (?q=) or filtered by tag (?tag=). When the harness’s search_memory tool calls GET /memory?q=..., the platform automatically bumps times_applied and last_applied_at on the returned rows so you can track which memories are actually being used.

​
Endpoints

MethodPathDescription
GET/api/v1/managed_agents/agents/{agent_id}/memoryList memories
POST/api/v1/managed_agents/agents/{agent_id}/memoryCreate a memory
PATCH/api/v1/managed_agents/agents/{agent_id}/memory/{memory_id}Update a memory
DELETE/api/v1/managed_agents/agents/{agent_id}/memory/{memory_id}Delete a memory

​
List memories

GET /api/v1/managed_agents/agents/{agent_id}/memory
Returns all memories for the agent. Use query parameters to narrow results. Query parameters
​
q
string
Case-insensitive substring search against memory text. When provided, only memories whose text contains this string are returned.
​
tag
string
Filter to memories whose tags array includes this exact tag value.
Both q and tag can be combined — results must satisfy both conditions. Returns an array of ApiMemory objects.

​
Create a memory

POST /api/v1/managed_agents/agents/{agent_id}/memory

​
Body parameters

​
text
string
required
The memory content to store. Must be at least one character.
​
tags
string[]
Up to 8 tags for categorizing and filtering this memory.
​
type
string
Optional free-form type label, e.g. preference, fact, instruction.
​
priority
number
Integer priority used to order memories when multiple are injected. Higher values are injected first.
​
source
string
Origin of the memory. One of: agent, slack, ui. Defaults to ui when called from the API. The harness sets agent; integrations like Slack set slack.
Returns the new ApiMemory object.

​
Update a memory

PATCH /api/v1/managed_agents/agents/{agent_id}/memory/{memory_id}
All fields are optional. Only fields you include are updated.
​
text
string
New memory content.
​
tags
string[]
New tag list. Replaces the existing tags.
​
type
string
New type label.
​
priority
number
New priority integer.
​
disabled
boolean
Set to true to suppress this memory without deleting it. A disabled memory is excluded from context injection and search_memory results but remains in the database for audit purposes.
Returns the updated ApiMemory object.

​
Delete a memory

DELETE /api/v1/managed_agents/agents/{agent_id}/memory/{memory_id}
Hard-deletes the memory row. Consider disabling (disabled: true) instead of deleting if you may want to restore the memory later. Returns 204 No Content.

​
ApiMemory response fields

​
id
string
Unique memory ID.
​
agent_id
string
ID of the agent this memory belongs to.
​
text
string
The memory content.
​
tags
string[]
List of tag strings.
​
type
string
Type label.
​
priority
number
Injection priority. Higher values are injected first.
​
disabled
boolean
When true, this memory is suppressed from context injection and search without being deleted.
​
times_applied
number
Number of times this memory has been returned by a GET /memory?q= search (i.e. how often the harness’s search_memory tool has fetched it).
​
last_applied_at
string | null
ISO 8601 timestamp of the last time this memory was returned in a search. null if never searched.
​
source
string
Origin of the memory: agent, slack, or ui.
​
created_at
string
ISO 8601 creation timestamp.

​
Example: create and list memories

curl -X POST "https://<your-platform-host>/api/v1/managed_agents/agents/agt_01j9xyz/memory" \
  -H "Authorization: Bearer $MASTER_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Always use TypeScript strict mode in this project.",
    "tags": ["typescript", "style"],
    "type": "instruction",
    "priority": 10
  }'

{
  "id": "mem_01n5xyz",
  "agent_id": "agt_01j9xyz",
  "text": "Always use TypeScript strict mode in this project.",
  "tags": ["typescript", "style"],
  "type": "instruction",
  "priority": 10,
  "disabled": false,
  "times_applied": 0,
  "last_applied_at": null,
  "source": "ui",
  "created_at": "2025-05-01T13:00:00.000Z"
}