CLI: Add agent subcommand for running AI agents

[rajeevramani]

Summary

Running agents currently requires 6 environment variables and knowing the Python script path:

FLOWPLANE_URL=http://localhost:8080 \
FLOWPLANE_TEAM=engineering \
FLOWPLANE_TOKEN=<token> \
LLM_BASE_URL=http://localhost:11434/v1 \
LLM_API_KEY=<key> \
LLM_MODEL=qwen3-coder-next:cloud \
python agents/dev_agent.py "Expose httpbin at localhost:8000 on path / at port 10001"

This should be a first-class CLI command that reads config from ~/.flowplane/config.toml.

Proposed Commands

# Run an agent (one-shot)
flowplane-cli agent run dev "Expose httpbin at localhost:8000 on path / at port 10001"

# Interactive mode
flowplane-cli agent run dev --interactive

# Streaming output
flowplane-cli agent run ops --stream "Why is /api/users returning 404?"

# List available agents
flowplane-cli agent list

# Verify a deployment (no LLM needed)
flowplane-cli agent verify --path /api/orders --port 10001

Config Extension

Add team and [llm] section to ~/.flowplane/config.toml:

token = "fp_pat_..."
base_url = "http://localhost:8080"
team = "engineering"

[llm]
base_url = "http://localhost:11434/v1"
api_key = "..."
model = "qwen3-coder-next:cloud"

Resolution priority (same pattern as existing config):

• team: --team flag > config file > FLOWPLANE_TEAM env var
• llm.base_url: --llm-url flag > config file > LLM_BASE_URL env var > https://api.openai.com/v1
• llm.api_key: --llm-key flag > config file > LLM_API_KEY env var
• llm.model: --llm-model flag > config file > LLM_MODEL env var > gpt-4o

Implementation Notes

• The CLI would shell out to the Python agents (or embed them via PyO3 later)
• agent list can discover agents from the agents/ directory
• config set needs new keys: team, llm.base_url, llm.api_key, llm.model
• Tested models: GPT-4o, Qwen 3 (via Ollama)

Milestone

Target: 0.0.15