Akasha¶

The shared cognitive fabric for intelligent agent systems.

What is Akasha?¶

Akasha is a distributed, real-time key-value fabric designed specifically for multi-agent AI systems. It gives your agents:

• Shared Memory — Working, Episodic, Semantic, and Procedural layers
• Stigmergy — Pheromone-based coordination (no message passing needed)
• Sub-millisecond latency — Faster than your LLM can think
• CAS — Compare-And-Swap for conflict-free concurrent writes
• Clustering — 3-node HA with CRDT + Raft consensus
• Observability — Prometheus metrics, Grafana dashboards

Quick Start¶

Get Akasha running in 30 seconds:

docker run -d -p 7777:7777 -p 50051:50051 alejandrosl/akasha:latest

pip install akasha-client

from akasha import AkashaHttpClient

client = AkashaHttpClient(base_url="https://localhost:7777")
client.put("agents/planner/state", {"status": "thinking"})

state = client.get("agents/planner/state")
print(state.value)  # {"status": "thinking"}

npm install akasha-memory

import { AkashaHttpClient } from 'akasha-memory';

const client = new AkashaHttpClient({
  baseUrl: 'https://localhost:7777',
});

await client.put('agents/planner/state', { status: 'thinking' });
const state = await client.get('agents/planner/state');

# Write
curl -sk https://localhost:7777/api/v1/records/agents/planner/state \
  -X POST -H "Content-Type: application/json" \
  -d '{"value": {"status": "thinking"}}'

# Read
curl -sk https://localhost:7777/api/v1/records/agents/planner/state

Why Akasha?¶

Most agent frameworks treat memory as an afterthought — a vector DB bolted on the side.

Akasha takes a fundamentally different approach: memory is the coordination layer itself. Agents don't need to talk to each other. They read and write to shared memory. They leave pheromone trails. They coordinate through the environment, not through messages.

This is stigmergy — the same mechanism ants use to build colonies without a central coordinator.

graph LR
    A[Agent Scout] -->|deposit pheromone| M[Akasha Memory]
    M -->|sense trail| B[Agent Analyst]
    B -->|write analysis| M
    M -->|read report| C[Agent Reporter]
    C -->|consolidate| M

    style M fill:#6c5ce7,stroke:#a29bfe,color:#fff
    style A fill:#00b894,stroke:#55efc4,color:#fff
    style B fill:#00cec9,stroke:#81ecec,color:#fff
    style C fill:#e84393,stroke:#fd79a8,color:#fff

Architecture¶

Next Steps¶