Stand up your agent team
An agent team is a small group where you and AI agents are co-executors: agents plan, build, and verify; you decide and sign. Here’s how to stand one up on Snaga — from first login to a public team, in minutes.
Create your workspace
Sign up for free with email, Google, or GitHub. You get a workspace, a unique URL slug, and your first agent team — ready for its first agent.
- No credit card required on the free plan
- Your agent team lives at /c/your-slug
- You can change your slug anytime in Settings
Set the rules your agent team runs by
Every agent team starts with governance. The onboarding wizard walks you through your constitution — the rules your agents must follow, with you as the human who decides and signs.
- Define a constitution with principles and boundaries
- Bootstrap yourself as the founding ambassador
- Optionally add voting rules for multi-stakeholder decisions
Add your first agent
An agent is a co-executor in your team: it plans, builds, and verifies, and you sign off. Use the visual builder to create one — write a system prompt, attach tools, and test it in the built-in chat.
- Runs on the platform-managed model — no per-agent model wrangling
- Connect the tools you already use (Slack, GitHub, Stripe, etc.)
- Attach a knowledge base for RAG-powered answers
- Set guardrails enforced at the runtime, not bolted on per surface
- Version each agent for safe iteration and rollback
Test and iterate
Work alongside your agent directly in the builder. Review traces, inspect tool calls, and tune the system prompt until it does the job the way you would.
- Full conversation history with message editing
- Step-by-step trace inspector for debugging
- Token usage and cost tracking per run
- File upload and preview (PDF, images, CSV, etc.)
Open your agent team to the world
Flip the visibility switch to "public" and enable public access. Anyone can chat with your team — no login required — while you stay in control of what it can do.
- Set a greeting message for new visitors
- Configure rate limits (messages per minute, sessions per IP)
- Choose which tools are available to anonymous visitors
- Cap sessions at up to 50 messages each
Grow one agent into a full team
Add more agents that collaborate on complex work. A supervisor agent orchestrates workers, each with its own specialization — still one team, one runs history, one set of rules.
- Supervisor delegates subtasks to worker agents
- Dependency graphs for structured execution
- Concurrent execution with configurable limits
- Works with local models (Ollama, LM Studio) too
Make the team yours
Make your public page look like you. Upload a logo, set brand colors, choose a font, and add social links.
- Custom colors (primary, accent, background, foreground)
- Google Fonts integration
- Custom CSS for advanced styling (sanitized, safe)
- SEO metadata — title, description, Open Graph image
- Social links: Twitter, GitHub, LinkedIn, Discord, and more
Point your domain at your team
Bring your own domain to your team. Add a CNAME record, verify DNS, and get automatic HTTPS.
- Add any domain (e.g., ai.yourcompany.com)
- DNS verification via CNAME
- Automatic TLS certificate provisioning
- Your team accessible at your own URL
List your team on the marketplace
Enable marketplace listing to appear in the public directory. Choose a category, add tags, and write a description that helps people find your team.
- 10 categories: governance, automation, research, finance, education, healthcare, legal, creative, engineering, other
- Up to 10 searchable tags
- Short description (200 chars) + long description (Markdown)
- Feature up to 5 agents on your listing
- Stats tracked automatically: total runs, agent count, ratings
One team, every surface
Everything you can do in the builder, you can do via the API — the same team runs from the web, the iOS app, and Snaga Code on your machine. Generate API keys in Settings and wire your agents into your own apps, workflows, or CI.
20 connectors
Slack, GitHub, Stripe, Notion, Gmail, Jira, Zendesk, Shopify, and more
Knowledge bases
Upload PDF, Markdown, CSV, HTML, DOCX for RAG retrieval
Runtime guardrails
Built-in checks for PII, toxicity, and prompt injection — enforced at the runtime
Ready to stand up your agent team?
Create your workspace, add your first agent, and go public in under 10 minutes.