Styrby vs. Dispatch: Remote Agent Control Compared

Dispatch is Anthropic's approach to remote agent control, designed to let developers manage Claude Code sessions from anywhere. Styrby solves a similar problem but for a broader set of agents. This article breaks down the architectural differences, security models, and practical tradeoffs so you can decide which fits your workflow.

Architecture: Platform Feature vs. Independent Layer

Dispatch is built into the Claude ecosystem. It uses Anthropic's infrastructure for session routing, authentication, and state management. This gives it tight integration with Claude Code's internals, including native access to the permission system, session context, and tool execution pipeline.

Styrby operates as an independent layer. The CLI connects to agents via their public interfaces and relays session data through Styrby's infrastructure. This means Styrby cannot reach as deep into any single agent as Dispatch can with Claude, but it can connect to agents that Dispatch does not support.

Agent Coverage

Dispatch works with Claude Code. That is its scope, and it does that well.

Styrby connects to five agents: Claude Code, Codex, Gemini CLI, OpenCode, and Aider. If you standardize on Claude Code exclusively, this difference does not matter. If your team uses multiple agents, or if you switch agents based on the task, Styrby's multi-agent support becomes meaningful.

Encryption Models

Dispatch relies on Anthropic's security infrastructure. Your session data is encrypted in transit (TLS) and at rest on Anthropic's servers. Anthropic can access the data if needed for support or compliance.

Styrby uses client-side TweetNaCl box encryption. Session messages are encrypted on your machine before they leave. The Styrby server stores only ciphertext. Even Styrby cannot read your session data.

The tradeoff is real. Anthropic's model enables features like server-side search across sessions and easier account recovery. Styrby's model provides stronger privacy guarantees but means that if you lose your device keys, your encrypted session history is unrecoverable.

Permission Handling

Dispatch inherits Claude Code's permission system natively. It has full visibility into what tools Claude is requesting and can enforce permissions with the same granularity as the local CLI. This is a genuine advantage of being a first-party tool.

Styrby intercepts permission requests at the CLI output level. It parses agent output to detect permission prompts and routes them to your mobile device. This works across all five agents but depends on parsing each agent's output format. When an agent changes its output format, Styrby's parsers need updating.

Cost Tracking

Dispatch does not include cost tracking or budget alerts. You track Claude costs through Anthropic's billing dashboard.

Styrby tracks token costs across all connected agents, updated on each page load. You can set budget alerts with graduated actions: notification at 80% of budget, slowdown at 90%, hard stop at 100%. For teams, costs are attributed per developer. Session tags let you label work by client or project for filtering.

Where Dispatch Wins

Deeper integration. Native access to Claude Code's internals means Dispatch can do things Styrby cannot, like manipulating the tool execution pipeline directly.
No additional cost. Dispatch is included with Claude Code.
Simpler setup. No separate CLI install, no account creation, no key management.

Where Styrby Wins

Multi-agent support. Five agents in one dashboard vs. Claude only.
Cost management. Per-agent and per-session tracking, budget alerts, session tags for client attribution.
Stronger encryption. Zero-knowledge E2E encryption vs. provider-managed encryption.
Vendor independence. Not locked into one AI provider's ecosystem.

The Broader Remote Control Landscape

Dispatch and Styrby are not the only approaches to remote agent control. The space is evolving quickly.

Codex. OpenAI runs Codex in a cloud sandbox, so remote access goes through the OpenAI dashboard. No mobile-native interface.
Gemini CLI and OpenCode. Neither offers built-in remote access. If you step away from the terminal, the agent waits.
Aider. Terminal-only. No remote monitoring. The git commit trail serves as an after-the-fact audit, but you cannot approve or deny actions remotely.
SSH/tmux workarounds. Some developers SSH into their workstation or use tmux to attach to agent sessions from a phone. This works but provides no cost tracking, no push notifications, and no permission classification.

The pattern is clear: each vendor is building remote access for its own agent. Nobody except Styrby is building cross-agent remote access. Whether that matters depends entirely on how many agents you use.

Making the Choice

If you use Claude Code exclusively and trust Anthropic with your session data, Dispatch is the simpler, cheaper option. If you use multiple agents, need cost controls, or require zero-knowledge encryption, Styrby addresses those requirements. The two are not mutually exclusive: you can use Dispatch for Claude-specific features and Styrby for cross-agent management.