Back to all stacks

Submit a Stack

Share your production AI agent stack with the community.

How Submissions Work

We're building a database of real production stacks, not theoretical combinations. Every submission goes through a review process:

  1. You submit details about your stack via email
  2. We verify the tools and architecture are real and compatible
  3. We may ask follow-up questions for clarity
  4. Approved stacks are published with "verified" or "reported" status

Verified means we've confirmed the stack is in active production use. Reported means it's been submitted but we haven't independently verified it yet.

What to Include

Stack Name & Pitch

A clear name and one-line description of what this stack does.

Use Cases

What problems does this stack solve? (e.g., customer support, coding assistant, research)

Full Layer Breakdown

  • • Orchestration: Framework/system coordinating the agent
  • • Memory: How state is persisted across conversations
  • • RAG: Vector store and retrieval setup
  • • MCP/Tools: External integrations and capabilities
  • • Eval/Tracing: Monitoring and improvement pipeline
  • • Models: LLMs powering the agent

Cost & Performance Notes

Rough estimates of costs and latency if known.

Caveats

Honest limitations, gotchas, or things you'd do differently.

Ready to Submit?

Send your stack details to our submissions inbox. Include as much detail as possible about each layer.

Email Your Submission

stacks@agentstackatlas.com

Alternative: Pull Request

If you're comfortable with Git, you can also submit a stack directly as a pull request. Each stack is a Markdown file with YAML frontmatter.

src/content/stacks/your-stack-name.md

Check the existing stack files for the schema format. We'll review and merge approved submissions.