Skip to content

About

Axonio SDKs: TypeScript (axonio), Python (axonio), MCP adapter (axonio-mcp). Persistent memory for AI agents.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

Axonio SDK

Documentation · MCP integrations · Retrieval benchmark

SDKs, CLI, ROS2 package and framework adapters for Axonio: persistent memory for AI agents. Formerly ClawDB.

The TypeScript, Python, Go and Rust packages are clients for the Axonio HTTP API. An API key alone connects them to the Axonio cloud; point them at a self-hosted clawdb-server with AXONIO_URL. The engine underneath that server lives in the repositories listed under The engine.

CI npm PyPI crates.io Go Founder School · FS26-2 alumni


Packages

Package Language Description
axonio TypeScript Client: HTTPS to the Axonio cloud, gRPC to a self-hosted server
axonio-cli TypeScript The clawdb command-line tool
axonio-langchain TypeScript LangChain.js adapter
axonio-openai-agents TypeScript OpenAI Agents SDK adapter
axonio-vercel-ai TypeScript Vercel AI SDK adapter
axonio-anthropic TypeScript Anthropic SDK adapter
axonio-google-genai TypeScript Google GenAI adapter
axonio-openclaw TypeScript OpenClaw adapter
axonio-mcp TypeScript MCP server (stdio transport)
axonio Python Sync and async client over HTTPS
axonio-ros2 Python ROS2 package: fleet-wide spatial memory for robots
axonio Rust Async client (tokio, rustls)
github.com/hackcat29/axonio-sdk/sdks/go Go Go client

The old names keep working: pip install clawdb and clawdb-ros2 install the new packages, the ClawDB class is still exported next to Axonio, and CLAWDB_* environment variables are read after AXONIO_*. The old @clawdb/* npm packages are deprecated with a pointer to these.

Architecture

There are two layers:

  • The engine (clawdb-server, written in Rust) is the database: storage, semantic search, branching, reflection.
  • The SDKs in this repository are clients that talk to a server, either the Axonio cloud or one you run yourself.

For a self-hosted server, one of these has to be true:

  • clawdb-server is already running locally. The HTTP API is on port 8080 and gRPC on http://localhost:50050
  • you ran npx axonio-cli@latest init, which provisions a local server
  • you are pointing the SDK at a managed cloud endpoint

The engine

The open-source engine is maintained in these repositories:

Repository What it is
hackcat29/ClawDB The runtime and clawdb-server binary
hackcat29/claw-vector Semantic memory engine: embeddings and vector search
hackcat29/claw-reflect Memory distillation: turns raw memories into durable summaries

They are forks of the original Claw-DB projects (now Axoniodev) and carry the fixes the Axonio cloud runs on: sessions, search, transactions and merge conflicts in the server; PostgreSQL migrations, UUID ids and workspace API keys in reflect; cluster-internal gRPC and a working embedder image in vector.


30-second quickstart

All of these use the Axonio cloud: create an API key in the dashboard and export it.

export AXONIO_API_KEY="ck_live_..."

TypeScript

import { Axonio } from 'axonio';

const db = Axonio.fromEnv(); // an API key alone means the Axonio cloud

const id = await db.rememberTyped('Deploy at 3 PM UTC', { type: 'task', tags: ['ops'] });
const hits = await db.search('deploy schedule', { topK: 5 });
console.log(hits[0]);

Python

from axonio import Axonio

with Axonio(api_key="ck_live_...", endpoint="https://api.axonio.xyz/api/v1/gateway") as db:
    memory_id = db.memory.remember("Deploy at 3 PM UTC", memory_type="task")
    hits = db.memory.search("deploy schedule", top_k=5)
    print(hits[0].content)

Rust

use axonio::{Axonio, RememberOptions, SearchOptions};

#[tokio::main]
async fn main() -> axonio::Result<()> {
    let db = Axonio::from_env()?;

    let id = db.remember_typed("Deploy at 3 PM UTC", RememberOptions::default()).await?;
    for hit in db.search("deploy schedule", SearchOptions::default()).await? {
        println!("[{:.2}] {}", hit.score, hit.content);
    }

    db.delete_memory(&id).await?;
    db.close().await;
    Ok(())
}

Go

package main

import (
	"context"
	"fmt"
	"log"

	axonio "github.com/hackcat29/axonio-sdk/sdks/go"
)

func main() {
	db, err := axonio.New() // AXONIO_API_KEY alone means the Axonio cloud
	if err != nil {
		log.Fatal(err)
	}
	defer db.Close()

	ctx := context.Background()
	id, _ := db.Memory.Remember(ctx, "Deploy at 3 PM UTC", nil)
	results, _ := db.Memory.Search(ctx, "deploy schedule", &axonio.SearchOptions{TopK: 5})
	fmt.Println(id, len(results))
}

For a self-hosted server, set AXONIO_URL (or pass an endpoint) and see each package's README.


Framework integrations

LangChain.js

import { ClawDBRetriever, ClawDBChatMessageHistory, createClawDBTools } from 'axonio-langchain';

const retriever = new ClawDBRetriever({ client: db, topK: 10 });
const history = new ClawDBChatMessageHistory({ client: db, sessionId: 'chat-123' });
const tools = createClawDBTools(db);

OpenAI Agents SDK

import { createClawDBAgentTools, ClawDBToolHandler, withClawDBMemory } from 'axonio-openai-agents';

const tools = createClawDBAgentTools(db, { enableBranching: true });
const agent = withClawDBMemory(myAgent, db);

Vercel AI SDK

import { clawdbTools, clawdbMiddleware } from 'axonio-vercel-ai';

const tools = clawdbTools(db);
const result = await generateText({ model, tools, prompt: 'What do you remember?' });

Claude Desktop (MCP)

{
  "mcpServers": {
    "axonio": {
      "command": "npx",
      "args": ["-y", "axonio-mcp"],
      "env": { "AXONIO_API_KEY": "ck_live_...", "AXONIO_AGENT_ID": "my-agent" }
    }
  }
}

CLI

npm install -g axonio-cli
clawdb init
clawdb memory search "deploy schedule"
clawdb branch fork my-experiment
clawdb sync
clawdb status

For ephemeral usage, prefer:

npx axonio-cli@latest init

Development

# Prerequisites: Node 20.14+, pnpm 9+
corepack enable
pnpm install

# Build all packages
pnpm -r build

# Run all TypeScript tests
pnpm -r test

# Run integration tests (requires running clawdb-server)
CLAWDB_INTEGRATION=1 pnpm --filter axonio-integration-tests test:integration

License

MIT — see LICENSE

About

Axonio SDKs: TypeScript (axonio), Python (axonio), MCP adapter (axonio-mcp). Persistent memory for AI agents.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages