# Koshra vs Rerun

> A factual comparison of Koshra and Rerun (rerun.build), from Rerun's public pages on 2 Nov 2026: licence, self-hosting, data residency, approvals, models and price, including where Koshra is behind.

Canonical: https://koshra.com/compare/rerun
Last updated: 2026-11-02

Rerun, from Reunit SA in Luxembourg, is a hosted service for AI coworkers, each with its own cloud computer, and since 29 October 2026 also an open source edition under AGPL-3.0. Koshra is a team of AI employees that runs on a server you control, built on an Apache 2.0 open source base. Both give a team of agents shared memory, schedules and approvals.

Compared on 2 November 2026 from the public pages of Rerun, linked below, with the model cards behind Koshra's free open-weight models.

| Area | Koshra | Rerun |
| --- | --- | --- |
| Where it runs | On your own server, in the country you choose; we set it up and support it with you | Hosted, with one VM per workspace on Hetzner in Germany. The open source edition can be self-hosted, which Rerun describes as documented, not supported |
| Licence | Built on an Apache 2.0 open source base: commercial use, changes and resale are allowed without publishing your changes. Commercial terms for Koshra: talk to us | Open source edition under AGPL-3.0, so a changed version you host for others must have its source published; its trademark policy asks anyone hosting it for others to rebrand |
| Data residency | Wherever your server is, for example India or the EU. Model calls go to the provider you pick for each bot | EU: workspace VMs in Germany, with US model providers among its listed sub-processors; GDPR data processing agreement |
| Team setup | Dispatcher with a plan card and countdown; one owner bot per task; create a bot from a description or from 48 role templates | A workspace assistant interviews you, plans a team and builds it; a board of coworkers |
| Approvals | Allow, Ask or Block per tool and space, approval cards, per-bot modes, CRM deletes blocked and bulk changes held | Approvals and questions with suggested answers; an autopilot answers coworkers within your rules, never for spending or deleting |
| Models | Model per bot, mid-chat switch, backup chain, cost on every reply; free open-weight models or your own key for OpenAI, Anthropic, Google, OpenRouter, MiniMax or any OpenAI-compatible endpoint | House tiers plus frontier models; your ChatGPT, Claude or Grok plan, API keys or local models |
| Apps and tools | Any MCP server, plus an in-app catalog of 2,137 apps through Composio and Nango, each allowed per bot | 3,000+ apps and MCPs on its site; a cloud desktop, browser and terminal per coworker that you can watch and take over |
| Uptime and speed | A public status page checked every minute; home page Lighthouse mobile performance 95 (our test, 2 Nov 2026) | No public status page found; home page Lighthouse mobile performance 40 (same test, same day) |
| For AI agents | Markdown for every page, llms.txt, a REST API with an OpenAPI 3.1 spec and an MCP server | llms.txt and Markdown on its docs site; an MCP API on paid plans |
| Devices | Web app on any browser, with phone alerts through Web Push | Web, an iOS app for its cloud, and a phone web app |
| Price | Talk to us; set with you around the team you need | Pro 29 USD a month for 5 coworkers and 500 credits, unlimited seats; 5-day trial with a card; open source edition free |

## Where Koshra is ahead

- Self-hosting is how Koshra runs, set up and supported with you, where Rerun's self-hosted edition is documented but not supported
- An Apache 2.0 open source base, so an agency can run it for clients under its own brand without publishing changes, where AGPL-3.0 and Rerun's trademark policy require published changes and a rebrand
- Data stays in the country you choose, including India, not only the EU
- A public status page, and a faster home page on the same Lighthouse test
- Rule-level approvals per tool and space, with CRM deletes blocked and bulk changes held
- Bots can run on free open-weight models with no model fees

## Where Koshra is behind

- Rerun is public with a published price and a self-serve trial; Koshra is in private preview with no self-serve sign-up
- Each Rerun coworker has its own cloud computer you can watch and take over; Koshra's visible browser is coming soon
- Rerun works with the ChatGPT, Claude and Grok plans people already pay for; plan sign-in for Koshra is coming soon
- Rerun has an iOS app; Koshra is a web app only
- Rerun publishes a GDPR data processing agreement; Koshra does not publish one yet and holds no certification
- Rerun has a template library and shares agents by link

## Coming to Koshra

Being built now, with no dates promised. None of this is counted above.

- Answers from the Brain that cite the notes they came from
- Building a whole team from one chat
- Simple model tiers, so you pick speed or depth instead of a model name
- Plan sign-in for ChatGPT and Claude
- A visible browser for each bot

## Questions

### Is Koshra an alternative to Rerun?

Yes, for teams that want AI employees on a server they control. Koshra is self-hosted and supported, built on an Apache 2.0 open source base, and keeps data in the country you choose. Rerun is a hosted service with an AGPL-3.0 open source edition.

### Can an agency resell Koshra but not Rerun?

Koshra's open source base is Apache 2.0, which allows commercial use and resale without publishing your changes; terms for Koshra are agreed with you. Rerun's open source edition is AGPL-3.0 and its trademark policy asks anyone hosting it for others to rebrand, so check both with your own counsel.

## Which to pick

If you want a hosted coworker with its own cloud computer today, a published price and a self-serve trial, Rerun is further along. If you want AI employees on a server you control, in the country you choose, under a licence that lets an agency resell, with approvals on every outbound step, Koshra fits.

## Sources

- [Rerun home](https://rerun.build)
- [Rerun pricing](https://rerun.build/pricing)
- [Rerun data processing agreement](https://rerun.build/dpa)
- [Rerun docs: open source vs cloud](https://docs.rerun.build/self-host/open-source-vs-cloud.md)
- [rerun-build/rerun on GitHub](https://github.com/rerun-build/rerun)
- [Rerun licence (AGPL-3.0)](https://github.com/rerun-build/rerun/blob/main/LICENSE)
- [Rerun trademark policy](https://github.com/rerun-build/rerun/blob/main/TRADEMARK.md)
- [NVIDIA Nemotron 3 Ultra model card (open weights)](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-Base-BF16)
- [dots3-note preview model card (Apache 2.0)](https://huggingface.co/dots-studio/dots3-note-prev)