← Back to the Hub

SuperQode CodeGem

codegem

Included

Google-focused six-tool coding harness for Gemini API and local Gemma, with optional local EmbeddingGemma code search.

PythoncodegemOpen sourceExact resume sessions

CodeGem tools run with host process permissions, without an OS sandbox or approval prompts.

Specification

CategorySuperQode harnesses
LanguagePython
How we knowRuns in the SuperQode engine (SuperagenticAI/superqode, Python).
Runtimecodegem
InterfaceTUI and CLI
Sourcebuilt-in
LicenceApache-2.0
ContinuityExact resume
Based oncodegem

Setup and support

Choose Local for Gemma or BYOK for Gemini. Gemini 4 Argon requires access on your Google API account. Tools use host process permissions.

Runtime behaviour

Capabilities

Gemini-native signed historyLocal Gemma through Ollama, LM Studio and llama.cppLocal EmbeddingGemma semantic searchResumable coding sessionsTerminal cancellation

Tools

readwriteeditsearchpythonbash

Policies and ownership

  • +Sandbox: none
  • +Approvals: none
  • +Shell: allowed
  • +Network: allowed
  • +Checks: not required

Get started

Inside SuperQode
$ :connect harness-codegem$ :harness show codegem$ :harness status
From your shell
$ superqode harness run codegem "Describe your task" --provider ollama --model gemma4:12b --working-dir .$ superqode harness run codegem "Describe your task" --provider google --model gemini-4-argon --working-dir .

Install and authenticate

  1. 1

    Install SuperQode

    uv tool install --python 3.13 --upgrade --force superqode
  2. 2

    Open CodeGem

    Enter :connect harness-codegem, then choose Local for Gemma or BYOK for Gemini.

    superqode

Evaluate and optimize

Measure
$ superqode harness test --spec harness.yaml$ superqode harness eval --spec harness.yaml --tasks eval-tasks.yaml$ superqode harness bench --manifest harnessbench.yaml
Improve
$ superqode harness optimize-omni --spec harness.yaml --tasks eval-tasks.yaml$ superqode harness optimize --spec harness.yaml --tasks eval-tasks.yaml$ superqode harness promote stage