Troubleshooting¶
This guide covers the most common SuperQode setup and runtime issues.
Start With Doctor Commands¶
Run the checks that match the feature you are using:
superqode doctor
superqode providers doctor
superqode runtime list
superqode harness doctor --spec harness.yaml
superqode sandbox doctor
superqode memory doctor
superqode trust doctor
Use JSON output when you need to paste diagnostics into an issue:
superqode doctor --json
superqode providers doctor openai --json
superqode harness doctor --spec harness.yaml --json
Installation Issues¶
ModuleNotFoundError: No module named 'superqode'¶
Reinstall SuperQode in the environment you are using:
python -m pip install --upgrade superqode
python -c "import superqode; print(superqode.__version__)"
With uv:
uv tool install --force superqode
superqode --version
For source installs:
git clone https://github.com/SuperagenticAI/superqode.git
cd superqode
python -m pip install -e .
The superqode command is not found¶
Check that the Python scripts directory is on PATH:
python -m site --user-base
python -m pip show superqode
If you installed with uv, check:
uv tool list
TUI Connection Problems¶
The TUI starts but no model is connected¶
Open the connection picker:
:connect
Or connect directly:
:connect byok openai gpt-4o-mini
:connect local ollama qwen3:8b
:connect acp opencode
Then confirm state:
:status
Provider authentication fails¶
Check the provider-specific environment variable:
echo "$OPENAI_API_KEY"
echo "$ANTHROPIC_API_KEY"
echo "$GOOGLE_API_KEY"
superqode auth check openai
superqode providers doctor openai
Never place API keys directly in documentation, issues, logs, or committed config. In superqode.yaml, reference environment variables:
providers:
openai:
api_key_env: OPENAI_API_KEY
ACP agent is unavailable¶
List and inspect agents:
superqode agents list
superqode agents doctor opencode
superqode agents doctor opencode --live
For OpenCode:
npm i -g opencode-ai
opencode --help
superqode connect acp opencode
Harness Problems¶
Harness YAML does not validate¶
Run:
superqode harness validate --spec harness.yaml
superqode harness inspect --spec harness.yaml
Common fixes:
- use
flavor: codingorflavor: no_tool - use a known runtime backend such as
builtin,openai-agents,adk,deepagents,pydanticai, orcodex-sdk - keep
execution_policy.allow_writeandexecution_policy.allow_shellfalse for no-tool harnesses - define
agentsas a YAML list
Optional runtime backend is missing¶
List backends:
superqode harness list-backends
superqode runtime doctor pydanticai
Install only the backend you need:
python -m pip install "superqode[pydanticai]"
python -m pip install "superqode[deepagents]"
python -m pip install "superqode[openai-agents]"
python -m pip install "superqode[adk]"
python -m pip install "superqode[codex-sdk]"
A run did something unexpected¶
Inspect persisted events:
superqode harness runs
superqode harness events <run-id>
superqode harness graph <run-id>
superqode harness graph <run-id> --json
Run doctor with the same runtime and sandbox overrides:
superqode harness doctor --spec harness.yaml --runtime pydanticai --sandbox local
Sandbox And Permission Problems¶
Shell commands are blocked¶
Check the active sandbox and harness policy:
superqode sandbox doctor
superqode harness inspect --spec harness.yaml
In the TUI:
:sandbox
:sandbox workspace-write
:status
If shell should be allowed, the harness must allow it:
execution_policy:
sandbox: local
approval_profile: balanced
allow_shell: true
allowed_commands:
- uv run pytest
A tool call is waiting for approval¶
In the TUI:
:approve
:approve 1 always
:reject
:reject 1 "use a safer command"
Use always only when you intentionally trust similar calls for the current session.
MCP Problems¶
MCP server is not available¶
Check config and dependencies:
superqode config init
node --version
npx --version
Example mcp_servers config:
mcp_servers:
filesystem:
transport: stdio
enabled: true
auto_connect: true
command: npx
args:
- -y
- "@modelcontextprotocol/server-filesystem"
- "."
Use mcpServers only when importing external MCP config. New SuperQode config should use mcp_servers.
Local Model Problems¶
Ollama model does not appear¶
Check Ollama:
ollama list
ollama serve
ollama pull qwen3:8b
superqode providers doctor ollama --live
Connect:
:connect local ollama qwen3:8b
DS4 is slow on first request¶
DS4 may warm up the model on connect. To skip warm-up:
export SUPERQODE_DS4_WARMUP=0
Then point SuperQode at the DS4 endpoint if needed:
export DS4_HOST=http://127.0.0.1:8000/v1
superqode providers guide ds4
Session And Export Problems¶
You cannot find an old session¶
superqode sessions list
superqode sessions tree
superqode sessions show <session-id>
You need to hand off a session¶
superqode share create <session-id>
superqode share list
superqode share import <artifact.superqode-share.json> --session-id imported
Export failed¶
Try a specific format and path:
superqode sessions export <session-id> --format markdown --output session.md
superqode sessions export <session-id> --format json --output session.json
Project Trust And Plugins¶
If plugins, hooks, or MCP config are blocked, inspect project trust:
superqode trust status
superqode trust doctor
superqode trust yes
superqode plugins doctor
Only trust projects whose local executable config you have reviewed.
Getting Help¶
When opening an issue, include:
- SuperQode version from
superqode --version - Python version from
python --version - operating system
- command you ran
- relevant
doctor --jsonoutput with secrets removed - minimal
superqode.yamlorharness.yamlthat reproduces the issue
Links: