Description
dcode is an open-source terminal coding agent designed for developers who need control over their coding environment. Built on the Deep Agents SDK, dcode allows users to customize their agent harness, control code execution, and bring their own models. It supports switching between models on-the-fly while retaining conversation state, enabling seamless transitions and continuity.
Read More
With dcode, users can share knowledge between sessions and agents using configurable memory, ensuring that valuable context is preserved across interactions. The platform offers Claude-compatible hooks and plugins for execution customization, allowing developers to tailor their coding agents to specific needs. Additionally, dcode supports adding custom models and providers through config files, providing flexibility in model selection.
dcode integrates with LangSmith for tracing and governing coding agents, offering insights into prompts, model responses, tool calls, errors, and token usage. Users can apply granular cost policy controls through LangSmith and LLM Gateway, ensuring efficient resource management. The platform provides secure sandboxes for agent tool execution, enhancing security and reliability.
Compatible with OpenAI, Anthropic, and OpenAI/Anthropic-compatible APIs, dcode allows users to switch providers or models based on quality, cost, and latency requirements. Model-specific prompting and tool descriptions can be automatically applied, optimizing agent behavior. Teams can share tuned configurations via config.toml, facilitating collaboration and consistency.
dcode is priced at USD 0, making it accessible for teams looking to ship great agents. It offers three ways to build, from full harness to full control, allowing users to choose the components they need and compose the rest of the stack around it. The platform is ideal for teams seeking an inspectable, configurable agent rather than a fixed coding assistant.
- CategoryAI Agent Builders
- Pricing ModelOpen Source
- DeploymentTerminal
- PlatformsWeb
dcode: Open-Source Terminal Coding Agent's Core Features
Open-source terminal coding agent
Customizable agent harness
Control code execution
Switch models on-the-fly
Configurable memory sharing
Claude-compatible hooks and plugins
LangSmith integration for tracing
Granular cost policy controls
How to use dcode: Open-Source Terminal Coding Agent?
Configure: Set up agent harness and models
Use: Execute code with custom models
Optimize: Apply model-specific prompts
Secure: Use sandboxes for tool execution
dcode: Open-Source Terminal Coding Agent's Use Cases
- Model Switching
- Custom Execution
- Token Management
- Secure Sandbox
- Team Collaboration
