At a glance
- What the listing says
- Open-source (Apache-2.0) platform from Parcha Labs, the team behind Grep, that adds learning between runs, durable recovery, reusable workflows and a management console to Claude Code, Codex, OpenCode and pi agents.
- Pricing
- Listed as Open source. For current plans, see the AgentRun site.
- Released
- Nov 10, 2026, per the vendor. Added to AgentStackd Nov 17, 2026.
- Category
- Automation & Agent Builders
Pros and cons
Pros
- Free to self-host under Apache-2.0 with your own model keys; start locally with npx @parcha/agentrun.
- Repeated steps can become reusable skills and workflows, which are checked and graded before later runs reuse them.
- Interrupted runs recover from the saved conversation, files and tool results instead of starting over.
Cons
- Launched as an open-source alpha.
- You pay your own model and infrastructure costs; the documented Render setup is about $27.50 a month before model, tool and computer usage.
- Learning adds its own reflection and grading cost, and savings depend on how much of your work repeats.
Sources, checked Nov 17, 2026: agentrun.ai/blog/introducing-agentrun-run-smarter-agents-use-fewer-tokens · agentrun.ai · github.com/Parcha-ai/agentrun. Plans and prices change, so confirm on the vendor's site.
Other Automation & Agent Builders
zapier.com
No-code automation across thousands of apps, with AI agents.
n8n.io
Fair-code workflow automation with AI agent nodes, self-hostable.
relevanceai.com
Platform to build and run teams of AI agents.
Featured on AgentStackd
Build AgentRun? Share this page or add the free badge to your site.
Is this your tool? Suggest an edit