OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Coding Agents

Dm Code Agent

Dm Code Agent is a local-first Python coding agent built around a ReAct loop. It works in a local project workspace to read and write files, run tests and lint, and call MCP tools, while recording plans, tool calls, and observations in append-only session logs. The repository also includes coding-agent evaluation workflows and extension points.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Dm Code Agent — OpenViST project profile in Coding Agents.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you want a Python coding agent whose work can be inspected after the run, or when you need a repository that includes reproducible coding-agent evaluation workflows.

Who it is for

  • Developers who want an agent to handle code-generation or code-editing tasks in a local workspace.
  • Developers who need session traces to diagnose, compare, replay, or branch coding runs.
  • Agent researchers who want a provided hidden-test benchmark or SWE-bench Verified evaluation workflow.
  • Python developers who want to extend an agent with tools, skills, providers, or lifecycle hooks.

Documented uses

  • Ask an agent to edit code and run tests or lint in a local workspace.
  • Inspect a run's plans, tool calls, and observations to investigate failures or skipped validation.
  • Compare two runs or branch a run from a recorded step.
  • Evaluate a coding agent with the included hidden-test tasks or the SWE-bench Verified workflow.
  • Add custom tools, skills, model providers, or lifecycle behavior through Python extensions.
Before you choose

Requirements & limits

The project is an open-source, MIT-licensed Python repository. Its quick-start coding workflow requires at least one API key. Tests, deterministic evaluations, and benchmark-manifest checks are documented as not requiring an API key; the browser read-only view can inspect sessions without one.

  • The primary coding workflow is CLI-based and operates in a local workspace.
  • Append-only JSONL traces support viewing, diagnosis, diffing, replay, and forking.
  • The browser workbench can submit tasks and show steps; read-only mode inspects sessions but cannot start runs.
  • Python 3.10 or later is listed as the requirement, with uv and pip installation paths documented.
  • Evaluation options include a 30-task hidden-test benchmark and SWE-bench Verified; the project says its older SWE-bench Lite subsystem was removed.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Dm Code Agent and configure a coding run

Set up the Python project and prepare the documented CLI workflow for a coding task.

Before you start

  • Python 3.10 or later
  • uv or pip
  • At least one API key for running a coding task

Steps

  1. Clone the repository and change into its project directory.

  2. Install dependencies with the command below or the command below.

    Documented command
    uv sync --frozen --extra dev
    Documented command
    pip install -e ".[dev]"

  3. Copy .env.example to .env and add at least one API key.

  4. Run dm-agent with a task prompt and --trace to save a session log.

Expected outcome

A coding task can be submitted through the CLI with its session recorded to a JSONL trace.

Guide limitations

  • The documented coding-task quick start requires an API key.
  • This setup recipe does not establish that a task was executed successfully.
Documentation 1 ↗
Source-based practical guide

Inspect and diagnose a recorded session

Use the trace CLI to review a session and examine reported failures or verification gaps.

Before you start

  • A session JSONL trace created by Dm Code Agent
  • The repository installed with its CLI tools available

Steps

  1. Run dm-agent-trace view with the session path to inspect its timeline.

  2. Run dm-agent-trace analyze with the session path to review the diagnosis.

  3. Use dm-agent-trace diff with two trace paths to compare runs, or dm-agent-trace fork with a trace and step ID to create a branch from that step.

Expected outcome

The trace can be reviewed, analyzed, compared with another run, or used as the starting point for a fork.

Guide limitations

  • Forking creates a new run path; it is not a guarantee that a task will succeed.
  • Replay is an explicit trace operation and should not be confused with a newly completed coding task.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for AI coding assistance

These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.

ProjectDocumented purposePlatformAccess and requirements
Dm Code AgentAsk an agent to edit code and run tests or lint in a local workspace.CLI, Self-hostedThe project is an open-source, MIT-licensed Python repository. Its quick-start coding workflow requires at least one API key. Tests, deterministic evaluations, and benchmark-manifest checks are documented as not requiring an API key; the browser read-only view can inspect sessions without one.
LlmcodeInspect a project's directory structure with `/tree`.CLIThe documented setup requires an API key and supports an API base URL and model setting. The project is open source under Apache-2.0.
Klaat CodeOpen a project directory for terminal-based coding assistance.CLI, ServerThe README provides a global npm installation command and documents browser sign-in with a KlaatAI account. It also identifies the repository as open source under Apache 2.0.
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
12 October 2026
Original website title
hwfengcs/DM-Code-Agent · Official homepage title · retained in its source language
How we review sourcesSuggest a correction