A lean coding assistant powered by OpenAI compatible inference endpoints LLMs and Serena MCP.
- 🚀 Powered by any OpenAI compatible endpoint
- 🛠️ Full Serena MCP tool integration
- ⚡ Fast compiled Go binary
- 🔧 Configuration management
go install github.com/unixsysdev/serena-cli-go/cmd/serena@latestCreate a serena-cli.yaml file (copy serena-cli.yaml.example from this repo). Config is loaded from:
./serena-cli.yaml~/.serena-cli/serena-cli.yaml~/.config/serena-cli/serena-cli.yaml
Example minimal config:
llm:
api_key: "your-api-key"
base_url: "https://llm.chutes.ai/v1"
model: "zai-org/GLM-4.7-TEE"
compaction_model: "Qwen/Qwen3-VL-235B-A22B-Instruct"
timeout_seconds: 300
serena:
command: "uvx"
tool_timeout_seconds: 300
enable_web_dashboard: false
enable_gui_log_window: false
max_tool_answer_chars: 20000
args:
- "--from"
- "git+https://github.com/oraios/serena"
- "serena"
- "start-mcp-server"Optional: set serena.context or serena.project_path if you want to force them;
leaving them empty lets Serena manage context and project activation.
Or set environment variables:
export LLM_API_KEY="your-api-key"
export LLM_BASE_URL="https://llm.chutes.ai/v1"
export LLM_MODEL="zai-org/GLM-4.7-TEE"
export LLM_COMPACTION_MODEL="Qwen/Qwen3-VL-235B-A22B-Instruct"
export LLM_TIMEOUT_SECONDS="300"
export SERENA_TOOL_TIMEOUT_SECONDS="300"
export SERENA_ENABLE_WEB_DASHBOARD="false"
export SERENA_ENABLE_GUI_LOG_WINDOW="false"
export SERENA_MAX_TOOL_ANSWER_CHARS="20000"# Run in current directory
serena
# Run in specific project directory
cd /path/to/project
serena
# Show configuration (API key is masked)
serena --config
# Show version
serena --version
# One-shot prompt
serena "summarize the repository"REPL commands:
/model
/model 3
/model "moonshotai/Kimi-K2-Instruct-0905"
/tools
/status
/context
/trace 5
/summary
/session list
/session new experiment
/session switch experiment
/compact
@context ./README.md
Tip: press Ctrl+C while a tool or model request is running to cancel it.
Sessions are stored under ~/.serena-cli/sessions/<project-name> so you can switch contexts.
On startup, the CLI prints a short summary of the session history (generated with the compaction model).
- deepseek-ai/DeepSeek-V3.2-Speciale-TEE
- MiniMaxAI/MiniMax-M2.1-TEE
- Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8-TEE
- moonshotai/Kimi-K2-Thinking-TEE
- moonshotai/Kimi-K2-Instruct-0905
- deepseek-ai/DeepSeek-V3.2-TEE
- zai-org/GLM-4.7-TEE
# Install dependencies
go mod download
# Run tests
go test ./...
# Build
go build -o bin/serena ./cmd/serena
# Run locally
go run ./cmd/serenaMIT