Prompt-defined AI Agents

Quick Example ​

Create a simple agent in seconds:

# agent.yaml
agent:
  name: "My Assistant"

llm:
  provider: openai
  model: "gpt-4o-mini"
  auth:
    api_key_env: "OPENAI_API_KEY"

mcp_servers:
  - name: "time"
    command:
      command: "uvx"
      args: ["mcp-server-time"]

Run it:

# Run the agent
gola --config agent.yaml

Then interact with the agent:

> What time is it?

GitHub-Native Agent Platform ​

Your agent is a repo. Share it like code.

# Run an agent directly from GitHub (fetches config and prompts)
gola --config github:username/sherpa-agent

# Specific version or branch
gola --config github:username/sherpa-agent@v1.2.0

# Custom configuration path
gola --config github:username/agents/customer-support.yaml

Fork, Improve, Share ​

# 1. Fork an existing agent
git fork github.com/gola-agent/gola-sherpa-agent

# 2. Customize for your needs
# Edit gola.yaml, add your prompts, configure tools

# 3. Share your improvements
git commit -m "Add corporate travel support"
git push origin main

# 4. Others can use your enhanced version
gola --config github:yourname/gola-sherpa-agent

Why This Matters ​

• One command - gola --config github:user/agent and you're running
• No package registry complexity - It's just a GitHub repo
• Version everything - Tags, branches, releases all work
• Collaborate naturally - PRs for prompt improvements? Yes
• Find agents easily - GitHub search already works
• See what you're running - Config and prompts are right there in the repo

What People Build ​

• Dev tools - Code review bots, doc generators, debugging assistants
• Content tools - Research agents, writing helpers, fact checkers
• Data tools - Query runners, report builders, analysis pipelines
• Ops tools - Deploy scripts, monitoring alerts, incident responders

Start Here ​

• Get Started - Build your first agent
• Configuration - All the knobs and dials explained
• Examples - Copy these and modify