Relayloop Cloud

An opt-in push target that adds durable off-machine backup and team features like pair check, with a web UI and cross-device pull on the roadmap.

Relayloop Cloud is a hosted layer on top of the local-first ai-hist CLI. The CLI stays exactly as it is — fully functional with no account, working offline. Cloud is an opt-in push target that adds what a local SQLite file on one laptop cannot: a durable off-machine copy of your history and team features built on a shared org corpus (today, pair check). A web UI, cross-device pull, a hosted MCP endpoint, and the Insights flywheel are on the roadmap.

The core promise is simple, and architectural rather than just policy: your data is yours to store and yours to remove at any time. See Privacy for the full data-handling model.

Why opt in

A local database is fast and private, but it lives on one disk. The moment you work across two machines, replace a laptop, or want a teammate to benefit from a session you ran, you hit its limits. Cloud closes those gaps.

Durable backup

Local files get corrupted or deleted. Pushing to the cloud keeps a durable copy of your record off your laptop.

Team recall

Once your team pushes to the same org, your collective history powers pair check — advisory warnings from past work before a risky change.

Cross-device access (coming soon)

Hydrate a new machine from your cloud history. Today push is one-way (local → cloud); pull is on the roadmap.

Web UI (coming soon)

A dashboard to search, browse sessions, and view Burn stats in the browser — no terminal required.

Hosted MCP (coming soon)

Point your Claude config at a hosted URL instead of running the local npx ai-hist-mcp server.

Insights (coming soon)

Cross-team patterns turned into proactive guidance and repo rules.

Opt in from the CLI

There is no ai-hist cloud subcommand. Cloud is two top-level commands: authenticate once, then push.

ai-hist login          # authenticate via Agent Relay Cloud; stores a token in $RELAYHISTORY_HOME
ai-hist push           # send new local history to the cloud (incremental, idempotent)

push is cursor-based — it only sends rows past the last accepted cursor, and the server deduplicates, so it is safe to run often (or on a loop alongside ai-hist watch). Exclude specific sessions with --incognito <session_id>. Your local commands — sync, search, show, session, stats — do not change.

Once your history is in the cloud, two team features build on it: ai-hist pair check surfaces advisory warnings from your team's past work before a risky change, and ai-hist learn distill turns sessions into structured decisions and findings. See the CLI reference.

Local vs. cloud at a glance

CapabilityLocal CLIRelayloop Cloud
Sync, search, show, statsYesYes
Works offlineYesLocal CLI still works offline
No account requiredYesAccount required
Durable off-machine backupNo (single disk)Yes (via push)
Team recall (pair check)NoYes
Cross-device pullNoComing soon
Web UI / Burn dashboardNoComing soon
Hosted MCP endpointLocal (npx ai-hist-mcp)Coming soon
Shared session management / InsightsNoComing soon (Teams)

The local CLI never breaks

This is the deal: nothing about going to cloud takes anything away locally. The CLI remains MIT-licensed and fully functional with zero cloud dependency. You can run entirely local forever, turn on cloud sync when you need backup or a second device, and turn on team features when your team is ready.

Start free and local. Turn on team sync — and the network — when you are ready.