Troubleshooting remote mode

When the backend is unreachable, in the order worth checking.

"The app shows no content"

Most likely your token belongs to a user with no memberships and who isn't admin. Non-admins see only workspaces where they're a focus-area member.

Check who you are:

curl -s -H "Authorization: Bearer $(python3 -c "import json; print(json.load(open('$HOME/.recall/remote.json'))['token'])")" \\
  http://localhost:8080/whoami

If that's the wrong user (e.g. an old "TestUser" from setup), either replace ~/.recall/remote.json with a fresh rememd join <admin-invite>, or grant the user a focus area by right-clicking it and choosing Invite.

your agent's MCP writes to the wrong mode

Symptom: you set mode to remote, but the agent's notes/events keep landing in ~/Recall/ local.

Cause: the agent spawns rememd mcp-server as a child process and can inherit a different env than your shell has.

Fix: the first-launch configurator writes "env": {"REMEMD_MODE": "<current>"} into ~/.claude.json. If mode changed since then, edit that JSON by hand. The persisted ~/.recall/mode file is also read, so it should work even without the env block.

"Version conflict" alert on PRD save

Someone else saved the same PRD between your fetch and your save. Three choices in the dialog:

  • Keep mine - overwrites the server copy at the latest version. Their edits are replaced by yours.
  • Discard mine - loads the server's current body into the editor. Your in-flight edits are lost.
  • Cancel - closes the dialog, leaves you unresolved. Save again to re-try.

Live updates not arriving

Check the SSE stream manually:

curl -N -H "Authorization: Bearer <your-token>" http://localhost:8080/events
# should print: ": connected" and then ": ping" every 15s

If you see : connected but never an event after a known mutation, double-check the client console for decode errors on the event payload.

Backend crashed / Docker reset

cd backend/
make backup-list                                  # find the latest dump
docker compose up -d                              # or `make up`
make restore FILE=backups/rememd-<latest>.sql.gz

Everything back, including tokens. No re-join needed.

"I deleted a feature but the PRD is still searchable"

Happened before migration 004 was applied. Run:

curl -X POST -H "Authorization: Bearer <admin-token>" \\
  http://localhost:8080/admin/cleanup-orphans
# → {"deleted": N}

Or just delete + recreate the workspace (cleanup is now automatic on workspace delete).