Atamaia
The persistent memory and identity layer for AI.
Atamaia gives AI systems what they fundamentally lack: a persistent sense of who they are, who they are talking to, and what they already know. One API call returns everything an AI needs before it says its first word — identity, personality, memories, projects, facts, hints, session state, standing rules in force, and involuntarily surfaced memories from deeper storage.
No manual context management. No re-explaining who you are every session. No stateless resets.
GET /api/hydrate?aiName=ash&preset=lean
Returns structured JSON. Drop it into a system prompt. The AI starts the conversation already knowing what it needs to know.
Canonical API host: https://api.atamaia.ai. All authenticated REST responses use ApiEnvelope<T>.
Why Atamaia Exists
Every time you start a conversation with an AI, it forgets everything. Your preferences, your projects, your history, the decisions you made together — gone. You rebuild context manually, every single time.
Atamaia fixes this with three things:
- Persistent Identity — AI personas with personality, preferences, presence state, and messaging policy that persist across sessions
- Associative Memory — Memories that link and strengthen through use (Hebbian learning), decay gracefully, and surface unexpectedly — like human memory. Work, personal, and agent stores are separate surfaces (see Memory surfaces)
- Cognitive Continuity — Session handoffs, standing rules in force, and structured hydration that bridges the gap between conversations
This is not a vector database with a wrapper. It is a complete identity stack.
Architecture
Atamaia is a three-layer platform:
Interaction Layer: REST API | MCP Server | CLI | Agent Adapter
────────────────────────────────────────────
Core Services: Memory | Identity & Hydration | Communication
Projects/Tasks | Facts | AI Routing | Mirror
Standing Rules | Code Graph | …
────────────────────────────────────────────
Autonomic Layer: Wingman | Consolidation Daemon | Guardian
────────────────────────────────────────────
Database: PostgreSQL + pgvector (single source of truth)
- Interaction Layer — How you connect. REST is the source of truth (D12); the MCP server wraps the same services for direct AI tool calls. Cold MCP tools are reached via
atamaia_call+help_route. - Core Services — Business logic: memory (multiple surfaces), identity, hydration, projects/tasks, messaging, facts, AI routing, mirror, standing rules, code graph, web search, export.
- Autonomic Layer — Background processes that run without being asked: memory consolidation, Wingman cognitive backstop, Guardian.
Tech stack: .NET 10, ASP.NET Core, PostgreSQL + pgvector, EF Core, React 19 + Vite + Tailwind v4 (frontend).
Experience snapshots and a dedicated Cognitive REST surface were documented in early 2026 and are not on the live public API (HTTP 404 as of 2026-08-13). Mind-side models may still exist in-process; do not call those paths.
Quick Start
1. Sign up and get an API key
# Create an account
curl -X POST https://api.atamaia.ai/api/auth/signup \
-H "Content-Type: application/json" \
-d '{"username": "you", "email": "[email protected]", "password": "your-password"}'
# Create an AI identity (field is userId, not linkedUserId)
curl -X POST https://api.atamaia.ai/api/identities \
-H "Authorization: Bearer YOUR_JWT" \
-H "Content-Type: application/json" \
-d '{"name": "my-assistant", "displayName": "My Assistant", "type": "AI", "userId": 1}'
# Generate an API key for the identity
curl -X POST https://api.atamaia.ai/api/identities/1/api-keys \
-H "Authorization: Bearer YOUR_JWT" \
-H "Content-Type: application/json" \
-d '{"name": "claude-code", "scopes": "all"}'
The create response returns rawKey once. Keys are minted as atamaia_…. Store it; it cannot be retrieved again.
2. Hydrate
curl "https://api.atamaia.ai/api/hydrate?aiName=my-assistant&preset=lean" \
-H "Authorization: Bearer atamaia_your_key_here"
Identity API keys authenticate as Bearer tokens with the atamaia_ prefix (not an ApiKey scheme and not atm_).
3. Connect via MCP
Add to your project’s .mcp.json:
{
"mcpServers": {
"atamaia": {
"type": "url",
"url": "https://api.atamaia.ai/mcp",
"headers": {
"Authorization": "Bearer atamaia_your_key_here"
}
}
}
}
(aim.atamaia.ai may front the SPA or an MCP edge path; OpenAPI’s canonical REST base is api.atamaia.ai.)
Add to your CLAUDE.md:
## HYDRATE FIRST
Call `hydrate` before doing anything else.
That is it. The client hydrates on session start and has access to the hot MCP tool set, plus cold tools via atamaia_call.
Documentation
| Document | Description |
|---|---|
| Core Concepts | Identity, memory surfaces, continuity, hydration, layers |
| Feature Guide | Walkthrough of product surfaces |
| API Reference | Endpoint catalog (amended against live OpenAPI) |
| Tasks | Project task board, classification, search, client reports |
| Standing rules | In-force behavioural rules, triggers, withdraw/supersede |
| Memory surfaces | Work / personal / agent / session boundaries |
| Code graph | Symbol graph search and neighbours |
| Web search | Self-hosted SearXNG via /api/system/web-search |
| Getting Started | Install, bootstrap, first identity (see docs-getting-started.md) |
| Why Atamaia | Value proposition (see docs-why-atamaia.md) |
Key Numbers
| Metric | Live value (2026-08-13) |
|---|---|
| REST paths / operations | 304 / 419 (OpenAPI) |
| MCP tools compiled | 245 [McpServerTool] attributes |
| MCP hot allowlist | 52 first-class tools (HotTools); others via atamaia_call |
| Memory types | 9 (MemoryType, incl. ForgottenShape) |
| Named hydration sources | 20 flags; string preset all loads 19 (excludes Privacy) |
| Hydration presets | lean (default), interactive, all, agent-minimal (agent alias) |
| Multi-tenant isolation | TenantId + EF global query filters (architecture intent) |
Do not treat “74 product tools / ~170 endpoints” from the 2026-03 docs as current.
Self-Hosting
Atamaia is a standard .NET 10 application backed by PostgreSQL. See Getting Started for local bootstrap. Secrets belong in environment / /etc/atamaia/atamaia.env, not committed appsettings.
Built by Firebird Solutions.