d:
GitHub
in development

d:spatch

Your agents. Your machines.
Your rules.

An open-source agent orchestration platform. Sandboxed execution. End-to-end encrypted. Control everything from any device.

Cross-platform
E2E encrypted
Sandboxed agents
Open source
build.log
✓end-to-end encrypted— your data never touches our servers
✓sandboxed execution— agents run in isolated containers
✓multi-device control— dispatch from phone, laptop, anywhere
✓bring your own agent— use any AI agent or framework
✓human-in-the-loop— built-in agent escalation system
✓open source core— free single-device tier, forever
◆awaiting dispatch
Get notified at launch
GitHub
© 2026 d:spatch
Features

Built for agents that actually ship

Everything you need to orchestrate AI agents across devices — with security that doesn't compromise on usability.

Sandboxed execution

Every agent runs as root in its own isolated environment — without compromising the host system. It can build, install, and execute anything it needs. Even docker.

Signal Protocol encryption

All user data is encrypted device-to-device with the same cryptographic industry standard behind Signal and WhatsApp. The server relays ciphertext it physically cannot decrypt.

Multi-device orchestration

Launch agents on your desktop from your phone — or from another desktop across the network. All commands are encrypted end-to-end. A device picker shows what's online and ready.

Human-in-the-loop

Easily have your agents escalate decisions with rich context and structured suggestions. They pause until you respond. Push notifications on every platform ensure nothing waits in silence.

Agent SDK

One async function, one context object in Python, Rust, or TypeScript. Report activity, stream logs, track token usage, and escalate to humans — all through a minimal SDK interface. Bring any framework and any model.

Open source core

The free tier is a complete product — local orchestration, Docker sandbox, and full dashboard. No server required, no account needed. The SaaS tier adds multi-device sync with encrypted communication.

Security

Zero-knowledge by design

All content is encrypted on-device before transmission using the Signal Protocol. The server never holds decryption keys — it forwards — and optionally stores — encrypted data it cannot read.

Your device
Plaintext in
End-to-end encrypted
Relay server
Ciphertext only
Encrypted relay
Their device
Plaintext out
Per-message keys
Every message is encrypted with a unique key. Compromising one reveals nothing else.
Forward secrecy
Encryption keys rotate automatically. Past messages stay secure even if a future key is compromised.
Trusted device network
Only devices you explicitly verify can join your network. No compromise.
Isolation

Every agent runs in a fortress

Agents run as root inside a sandboxed container with full system access — including their own Docker daemon. They can build images, run sub-containers, install packages, and execute arbitrary code. Your host machine stays completely sealed off.

Inside the sandbox
Agent container
✓Running as root
✓Workspace mounted
✓Full system access
✗No host access
Nested Docker daemon
build · run · compose
Agent SDK & Hub

Ship agents in minutes

The Agent SDK gives your agent a single context object. Plug in Claude Code, define your tools, and let it run — sandboxed, with all permissions bypassed. The d:spatch Agent Hub makes it easy to share your agents with the community, and integrate them into your workflows.

agent.py
from dspatch import SDK
from claude_agent_sdk import ClaudeAgentOptions, ClaudeSDKClient, ResultMessage

sdk = SDK()

@sdk.agent
async def coding_agent(prompt: str, ctx: sdk.Context) -> str:
    options = ClaudeAgentOptions(
        system_prompt="You are an autonomous coding agent.",
        mcp_servers={"dspatch": ctx.mcp_server},
        allowed_tools=["Read", "Write", "Edit", "Bash", "mcp__dspatch__send_inquiry"],
        permission_mode="bypassPermissions",
        cwd=ctx.project_path,
    )

    async with ClaudeSDKClient(options) as client:
        await client.query(prompt)
        async for msg in client.receive_response():
            if isinstance(msg, ResultMessage):
                return msg.result

sdk.run()
Platforms

One app. Every device.

A single codebase runs on all five platforms. Desktop devices orchestrate agents locally. Mobile devices control desktops remotely — all through the same encrypted channel.

Desktop
WindowsmacOSLinux

Run agents locally with full Docker access. Build, test, and deploy from any desktop.

Desktop & Mobile
WindowsmacOSLinuxiOSAndroid

Control any desktop remotely through the encrypted relay. Monitor, respond to inquiries, and launch agents on the go.

Early access

Be first to dispatch

d:spatch is in active development. Join the waitlist to get early access when we launch.

Get notified at launch
Open source release·Desktop apps·Mobile apps·Python SDK