🎉 We just closed a $1M Private Round!Read the announcement →

One agent or a whole fleet — memory that scales.

Fleet Overview
Collective Memory
Version 2.0

d33pmemory uses named agents and optional teams to scope memory. Each agent builds its own private knowledge. Agents in the same team share a collective memory pool — new agents instantly inherit everything the team has learned.

How it works

Step 01

One API key, many agents

Your API key belongs to your account — not to a single agent. Create as many named agents as your plan allows. Each agent (e.g. 'slack-bot', 'personal-assistant') gets its own isolated memory.

Step 02

Agents have names, not keys

Pass agent_id in every ingest and recall call. d33pmemory routes the memory to the right agent automatically. No per-agent API keys needed.

Step 03

Teams = shared memory pools

Assign agents to a team and they share a collective memory pool. One agent learns something — every agent on the team benefits. New agents join and immediately have access to everything the team has ever learned.

Step 04

Private by default

Memories extracted from an agent's conversations are private to that agent unless the extraction engine classifies them as team-relevant. No cross-contamination.

Step 05

Instant inheritance

Plug in a new agent, assign it to a team, and it immediately has access to the full shared pool. Zero cold start.

Architecture

Your account (1 API key)

├── agent: slack-bot

└── private memories

├── agent: personal-assistant

└── private memories

└── team: customer-support

├── agent: support-bot-1 → private + shared

├── agent: support-bot-2 → private + shared

└── shared pool → all team agents read/write

Ingest — pass agent_id
POST /v1/ingest
{
  "agent_id": "personal-assistant",
  "user_message": "Book me a table for 2",
  "agent_response": "Booked at Trattoria, 8pm"}
Recall — scoped to agent + team
POST /v1/recall
{
  "agent_id": "personal-assistant",
  "query": "dining preferences"}
Memory scopes
"own"Only this agent's private memories
"shared"Team shared pool only
"all"Private + team shared combined
"auto"Own first, include shared if needed — recommended
Setup

Getting started

01

Create an agent

POST /v1/agents — give it a name like 'personal-assistant' or 'slack-bot'. That name is your agent_id in every API call.

02

(Optional) Create a team

POST /v1/teams — if you want agents to share memory. Assign agents to the team via POST /v1/agents/:id with a team_id.

03

Ingest

After every conversation, POST /v1/ingest with agent_id. d33pmemory extracts and stores memories under that agent.

04

Recall

POST /v1/recall with agent_id and a natural language query. Get back ranked memories, ready to inject into context.

Plans

How many agents & teams?

FeatureBasic $19/moPro $39/mo
Agents3Unlimited
Memories stored5,000100,000
Ingests/month2,00020,000
Teams3Unlimited
Extraction + Recall
Auto consolidation
Fleet/collective memory
Recall scoping
Priority support

Full plan comparison →

Build your first agent

One API key. As many named agents as your plan allows.