Runsight

What is Runsight?

Runsight is an open-source, YAML-first workflow engine for AI agents. Engineers design agent workflows on a visual canvas while clean YAML is written directly to the filesystem. That means workflows can be committed, diffed, and reviewed in pull requests, giving production teams real version control and rigor.

Top Features:

• YAML-first workflows: stored as Git-native files with no platform vendor lock-in.
• Canvas plus editor: visual workflow design stays synced with the underlying YAML file.
• Per-run cost tracking: shows real token spend down to the cent with budget caps.
• Eval framework: built-in assertions and regression tests catch agent drift early.
• Mid-flight controls: pause, fix, or kill running agents without redeploying a service.

Use Cases:

• Production agents: run business-critical AI agents with audits, rollbacks, and pull request reviews.
• Agent experiments: iterate rapidly on prompts and tools while tracking cost per run.
• Regression testing: lock in agent behavior using assertions and automated evaluation suites.

Who Can Use Runsight?

• Platform engineers: host reliable agent runtimes with cost caps and clear operational observability.
• AI developers: design, version, and debug multi-step agent workflows in a familiar Git flow.
• Startups: ship agent features fast while keeping compliance-friendly code review in place.

Pricing

• Open source (free): full source released under Apache 2.0 with self-hosting on your own infrastructure.
• Bring your own keys: pay only the underlying model provider for agent token usage.
• Self-hosted: run locally or inside private infrastructure without recurring platform license fees.

Pros and Cons

Pros:

• Git-native: workflows live as YAML files reviewed through normal pull request flows.
• Open source: Apache 2.0 license permits commercial use and deep customization.
• Operational controls: cost caps, pause, resume, and eval framework fit production workloads.

Cons:

• Self-hosting burden: teams must run, patch, and monitor the server themselves.
• Learning curve: YAML-first design requires engineers comfortable with configuration-as-code patterns.
• Young project: ecosystem and integrations are still growing compared to larger agent platforms.

FAQs:

1) Is Runsight free?

Yes, Runsight is open source under Apache 2.0 and self-hosted on your own servers.

2) Do I need to write YAML by hand?

No, the visual canvas writes clean YAML, but you can edit it directly if preferred.

3) Can I pause a running agent?

Yes, you can pause, fix the prompt, and resume the agent with no redeployment.

4) Which model providers are supported?

Runsight is model agnostic and works with any provider using your own API keys.

5) Is there a hosted version?

Runsight is primarily self-hosted, so teams run it inside their own private infrastructure.