Gravel
allow non-engineers to make prompt PRs
Last verified:
What is Gravel?
Gravel is an open-source library that embeds a prompt management dashboard directly into your existing AI application. It allows AI engineering teams to manage prompts, trace outputs, and create evaluations without leaving their codebase. The tool is installed into TypeScript or Python codebases and mounts an admin dashboard inside your app.
Key features include embedded prompt editing that lets non-engineers (like lawyers, clinicians, and domain experts) make prompt changes through PRs without knowing what a PR is. It provides production trace review capabilities where team members can review actual outputs from production. Reviews automatically create prompt change suggestions and evals. The tool maintains a single source of truth for prompts in your codebase using git version control.
Gravel is designed for startups and teams building vertical agents who need to decouple engineer work from domain expert alignment. Engineers build agents with git while non-engineers align agents outside of git through the embedded dashboard. It supports observability through comprehensive tracing, error analysis, and searchable traces with inputs, outputs, and metadata.
The platform is open-source under Apache 2.0 license and is part of the Artanis observability platform ecosystem. It solves the coordination and versioning problems that occur when using separate docs and platforms for prompt management.
Gravel pricing
Pricing model: Freemium
Free and open-source under Apache 2.0 license. No paid tiers mentioned - the library is completely free to use and modify.
Gravel pros
- Open-source under Apache 2.0 license
- Embeds prompt editor directly into your app
- Single source of truth in codebase with git
- Non-engineers can make prompt PRs without knowing Git
- Review production outputs straight from traces
- Reviews automatically create prompt change suggestions
- Reviews automatically create evals
- Mounts admin dashboard inside your existing app
- Supports both TypeScript and Python
- Works for vertical agents and domain experts
- Domain experts (lawyers, clinicians) can align agents
- No separate platform or docs to maintain
- Solves coordination and versioning problems
- Part of Artanis observability platform ecosystem
- Sets up in about a minute
Gravel cons
- Requires installing library into existing codebase
- Only supports TypeScript and Python
- Needs git workflow already in place
- May require engineering setup time initially
- Limited to teams already using git
- Dashboard mounts inside your app (not standalone)
- New project with potentially limited documentation
- Requires AI engineering codebase to work with
Frequently asked questions about Gravel
What is Gravel?
Gravel is an open-source library that embeds a prompt editor straight into your AI application, allowing teams to manage prompts, trace outputs, and create evaluations with a single source of truth in your codebase.
Who is Gravel for?
Gravel is for AI engineering teams building vertical agents, particularly startups where engineers build agents with git and non-engineers (lawyers, clinicians, domain experts) need to align agents outside of git.
How do non-engineers use Gravel?
Non-engineers can make prompt PRs without knowing what a PR is. They review outputs from production traces and their reviews automatically create prompt change suggestions and evals through the embedded dashboard.
What programming languages does Gravel support?
Gravel supports TypeScript and Python. You install the library into your existing AI engineering codebase in either of these languages.
Is Gravel open-source?
Yes, Gravel is open-source under the Apache 2.0 license, which means it's free to use, modify, and distribute.
How does Gravel handle version control?
Gravel maintains prompts with a single source of truth in your codebase using git. This solves coordination and versioning problems that occur with separate docs and platforms.
How long does setup take?
Setup takes about a minute. After installation, your team can immediately start making prompt PRs and reviewing production outputs.
What happens when non-engineers review outputs?
When non-engineers review outputs from production traces, those reviews automatically create prompt change suggestions and evals that can be used to improve the AI system.
Does Gravel work standalone or with other tools?
Gravel is a library you install into your existing AI application. The admin dashboard mounts inside your app rather than being a standalone tool, integrating with your existing codebase.
What is the relationship between Gravel and Artanis?
Gravel is part of the Artanis ecosystem. Artanis is an observability platform for AI applications that provides comprehensive tracing, and Gravel specifically handles embedded prompt management, tracing, and evals for AI engineering teams.