Connector setup and field reference

After saving, use the connector’s test or General → Run health check. Refresh Settings and the sidebar if their snapshots differ.

In Settings → Connectors → Jira, enter your token and click Save credentials or Save token. For Jira Cloud, supply the Atlassian account email that owns the API token. A GHE token or Google login cannot authenticate Jira.

Test by opening a ticket you can access and running health checks. A saved token does not guarantee access to every project. Ticket transitions and comments follow workflow approvals and your provider permissions.

1. Open P4V and identify the exact server, user and workspace/client you use for the ticket’s depot files. Ensure required VPN/network access.
2. In Fabric’s Perforce section, click Auto-detect or enter the actual p4 executable. P4V alone is not proof that the CLI is installed. Auto-detect does not install or save anything.
3. Copy the server, user, client and exact client root into the fields below. Click Save & test.
4. Resolve authentication or trust through your normal Perforce login process. Then test again. Use the same CLI/server/user identity that Fabric uses.
5. If a depot file is unmapped, inspect the selected client’s view and p4 where result with your maintainer. Do not guess depot paths or change to a different client just to silence a warning.

Root mismatch example: a folder can exist and show “found” in General while still being the wrong P4 client root. Correct both General → Perforce workspace and this connector’s root. The Unity project and Live-data Git checkout are separate folders; they need not live under Perforce.

1. Install/open the Photoshop version your workflow supports. Open Settings → Connectors → Photoshop.
2. Use Show plugin manifest… to locate the packaged Fabric Intelligence Provider. Load its manifest through Adobe UXP Developer Tool following the supplied plugin instructions.
3. Use the pairing controls in Settings and the provider to connect to this Executor. Keep Photoshop and the provider running.
4. Confirm paired and ready / UXP bridge connected, not just “app ready”. A preparation stage also needs the correct source document and permitted operation.

Plugin source points into the installed release’s resources. Re-pair… replaces the local pairing credential and disconnects sessions using the old credential. After an upgrade, verify that the loaded plugin points to the intended installed version. Pairing secrets stay on this Mac.

1. Set General → Unity project and Unity executable to the required project/editor. Open that project in Unity and wait for it to finish loading.
2. Start the project’s supported Unity-MCP bridge. Use its actual host and port in Settings → Connectors → Unity.
3. Enter a loopback host (normally 127.0.0.1) and the bridge port (often 8765), then Save & test.
4. If connection is refused, check whether the bridge is running in that editor and whether the port matches. The Fabric dashboard port is a different service.

Fabric accepts loopback hosts for this connector. The presence of a Unity installation or project does not establish a bridge connection.

Clone the required repositories with your normal Git client and configure the correct Git user identity and remotes. In Connectors → GitHub (push), save your GHE PAT if your deployment uses it, then check the Git credential helper and live health. Network access to your organization’s Git host is still required.

Git worktree and Live-data repository select local repositories; a PAT does not clone them. A readable clone does not prove fetch/push permission. Fabric’s governed publisher prepares the ticket branch and exact publication plan. Do not commit game assets directly to main to work around a blocked run. Delivery and PR process.

1. Obtain a registered Google Desktop OAuth client JSON from your maintainer or a Google Cloud project you are authorized to manage. Its project needs Sheets API enabled and consent configured.
2. Expand Google sign-in setup and choose the client JSON (Desktop installed-app format, under 64 KB). A service-account JSON or a text file containing an email is not this client.
3. Click Sign in with Google and complete consent on Google’s browser page with your own account. If the browser did not open, use the displayed continuation link.
4. Click Refresh status. Check whether the saved grant is read-only or read/write. Reconnect if the required grant changes.
5. Confirm your Google account can open the exact sheet and tab the workflow needs. Ask the owner to share it if needed. Configure RAG sources or the workflow’s sheet targets separately.

Why is sign-in disabled? No registered Desktop OAuth client has been imported, or configuration is still loading/busy. Fabric cannot enable Google login by storing your Google password. No client available means OAuth setup remains blocked.

Content Bundles and the shared Jira completion ledger are workflow targets, not automatically all sheets in your Drive. Ask your workflow owner for the exact sheet IDs/tabs and access requirements. Adding a Sheet as knowledge does not authorize writes; publishing rows still requires write scope and the exact workflow approval.

Save your Confluence PAT, then click Test page access. Check the configured site and space in General → Integrations. A token marked saved can still lack page visibility. Verify the page URL, account permissions and VPN. Add relevant page URLs under Knowledge as Web URL sources, then rebuild; authenticated ingestion uses the backend credential, not your browser cookies.

Issue evidence is input to investigation. A readable issue or discovered tool does not authorize a repair, deployment or outward message.

Use the team-authorized bot token in the masked token field. The workspace panel supports labeled access tokens and an optional refresh token. Connect only workspaces and recipients the workflow is allowed to use. Incoming-webhook configuration, where supplied by the deployment, is scoped to that webhook’s destination.

The Test DM to field is an email or handle. Sending a test is an actual outward message; use an agreed test recipient and the required confirmation. A successful connection never grants blanket messaging permission.

This connector reports the inference runtime. Configure providers, models, keys and routing under Settings → AI · Models. Keep Ollama/Fabric Local AI available for the local capabilities in use, including embedding. A chat model being available does not prove that the selected vision model or embedder is ready. AI field reference.

1. Obtain the running MCP service endpoint from its provider. A GitHub repository URL is source code, not an endpoint.
2. Enter Name, Endpoint and optional Notes. Select No runtime capability unless this provider genuinely supplies the specified Browser Control or Computer Control capability.
3. Click Add, discover its tools, and review the exact tools to allow. A registered endpoint is not yet executable authority.
4. Bind approved tools to the exact supported runtime node/capability. Test only an appropriate, bounded operation; provider credentials and action approvals still apply.
5. If discovery fails, verify the provider is running, the endpoint protocol is supported and credentials are available. Do not mark the connector connected based only on reachability.

Disabling a connector removes it from availability. A declaration, a successful tools list or a downloaded skill does not certify the provider or authorize arbitrary tool calls.