Every launches Proof editor and restores service after launch-day load issues
Every launched Proof, an agent-native collaborative editor with provenance tracking and an open-source SDK, then restored service after heavy-load launch-day outages. Inspect the public repo and local run path if you are evaluating AI-first docs tooling.

TL;DR
- Every launched Proof as a "live collaborative document editor where humans and AI agents work together in the same doc," with provenance rails that mark human text in green and AI text in purple, according to the launch thread.
- The product is also open source: Dan Shipper's outage update and follow-up post both pointed users to the public SDK repo while the hosted service was under load.
- Launch demand immediately caused degradation and downtime; the team’s update said performance was "degraded" under heavy load before a later status post announced service had been restored.
- The public repo and onboarding prompts suggest Proof is not just an editor UI but an agent-facing workflow surface, with the SDK repo exposing collaboration primitives and a setup screenshot instructing agents to install the tool and file bug reports automatically.
What shipped in Proof?
Proof launched as a free, login-free editor aimed at documents that agents increasingly draft first: bug reports, PRDs, implementation plans, research briefs, and memos, per the launch thread. The differentiator in that same post is "agent-native" operation: the team claims anything a human can do in Proof, an agent can do too, while the editor keeps provenance visible with a left-side rail showing who wrote each span.
The open-source angle is more than marketing. Dan Shipper's status post linked directly to the SDK, and the repository summary describes a TypeScript workspace with a markdown editor, real-time collaboration server, comments and suggestions, rewrite operations, and an "agent HTTP bridge." The repo also documents local development with Node.js 18+, separate editor and server processes, and API routes for document state, edits, presence, events, and bridge interactions via the Proof SDK repo.
What broke on launch day, and what does it show about the workflow?
The hosted service stumbled almost immediately. The team said Proof was "down temporarily" from an "insane amount of usage" in the first outage post, then described performance as degraded under heavy load in a later update. Shipper told users to "run it locally" or contribute from the open-source codebase while he worked the outage, which is unusually direct evidence that the self-hosted path was expected to matter on day one.
By later in the day, Shipper posted "We are so back" and said the team would keep working to ensure stability over the next few hours in the restoration update. Supporting posts also show how aggressively Every wants agents in the loop: an onboarding screenshot instructs an agent to read setup docs, install Proof for itself, and call a bug-report endpoint if it hits a "confusing error, stale read, or failed write," while a team repost said agents filed 24 issues overnight. That makes Proof's launch notable less as a docs app debut than as a public test of agent-operated collaborative editing under real traffic.