Start guides
Start
Troubleshooting
Recover from common setup, access, search, and action failures without guessing.
Updated September 5, 2026 3 min read
Start with the symptom you can observe. Record the time, workspace, page, and exact error before changing configuration. Keep credentials and customer message content out of shared reports.
The local app will not start
No package.json or npm cannot find the project: run commands in the cloned repository root, where package.json lives. A parent folder containing several repositories is not the application directory.
Unsupported Node version: run node --version; the project requires Node.js 22 or later. Install a supported version using your development environment's normal setup, then run npm ci in the project.
EADDRINUSE or port already in use: check whether your own development server is already running. Reuse it or stop it with Control-C in its terminal. To choose another port, use npm run dev -- --port 3001. For a connected installation, also update the app origin and authentication callback configuration. Do not force-quit unrelated processes.
Resource gate refuses a build: retain the refusal, close only your own disposable jobs, and use CI if the machine lacks capacity. Do not raise limits or start another build in a different worktree.
Sign-in works but the workspace will not open
Confirm you are using the intended installation and account. A user needs an active membership and a usable workspace. Ask the platform owner to inspect the membership and workspace status.
For a fresh self-hosted installation, check that ADMIN_EMAIL and BOOTSTRAP_FOUNDER_EMAIL match and that the owner existed before migrations. Follow the installation sequence. Re-running an already recorded migration is not a membership repair.
A feature is missing
Check your current business context, then the enabled modules in Integrations. A disabled module may remove navigation and refuse direct access. Provider-dependent features also need a configured, verified connection. Use Setup Center and the capability reference.
The fictional demo does not establish that every capability is configured in a connected workspace.
An invitation or message has an uncertain result
Read the exact warning and inspect the activity or provider receipt. A draft is not sent; approval is not delivery; provider acceptance is not proof of inbox arrival.
If the result is unknown, check the provider and local activity before retrying. Repeating an action with a new request can create a second external effect. For an explicit invitation delivery failure, use the directory's supported resend flow after correcting the connection. See action results.
An import contains duplicates or the wrong people
Pause further imports. Review the file's mapped columns and ambiguous matches using Importing contacts. Keep a copy of the original file in your approved storage. Do not delete records in bulk to repair an unexplained mapping problem.
Search cannot find the guide
Search for the task, such as “invitation”, “first developer change”, or “migration”. Exact AI tool names also work. Clear the search to browse sections. If search reports a loading error, use Retry; the section navigation remains available.
Give someone enough information to help
For your technical owner or a sanitized repository issue, include:
- What you wanted to do and the smallest repeatable steps.
- The installed commit or version, affected route, browser, and time of failure.
- Whether it happened in the fictional demo, a local test workspace, or a deployed workspace.
- Expected behavior, actual behavior, and the error or receipt identifier.
- Which verification failed and whether anything external may already have happened.
Remove tokens, environment files, invitation links, customer data, and raw message bodies. Route production incidents to the agreed operator; the repository issue tracker is not a promised emergency support channel.