> ## Documentation Index > Fetch the complete documentation index at: https://docs.get-hive.ai/llms.txt > Use this file to discover all available pages before exploring further. # Troubleshooting > Fix the common problems in Hive: connector errors, an empty Signals feed, workflow deploy issues, public forms, billing restrictions and halted actions. Start with what Hive shows you. The status on a connector card, the message on a deploy, the banner at the top of the screen and the [Audit log](/approvals/audit-log) usually point straight at the layer that needs attention. ## Connectors and integrations ### A connector says Reconnect required or Error The provider's token has expired or its permissions were withdrawn. Go to **Integrations**, find the tile and open its **Accounts** panel. Sign in to the provider again and approve the permissions it asks for. Approving fewer permissions than requested is the most common reason a connection fails again. An Admin can select **Run loop now** on **Signals** to sync immediately instead of waiting for the hourly loop. A connection shown as **Needs reconnect to establish an owner** predates personal account ownership. Reconnect it and it becomes yours. ### "A workspace administrator must enable this provider" Members can connect their own accounts only after an Owner or Admin has enabled that provider for the workspace. Ask an Admin to connect it once — that enables it for everyone. See [Accounts and autonomy](/integrations/accounts-and-autonomy). ### PostHog connects but returns no data PostHog asks you to choose a region when you connect: **European Union (eu.posthog.com)** or **United States (us.posthog.com)**. The wrong region resolves to a different host, so reads come back empty. Disconnect and connect again with the region your PostHog project lives in. ### Sage 200 refuses the connection Choose the edition you use — **Sage 200 Standard Online** or **Sage 200 Professional / Extra Online** — when asked. The Sage user you sign in with needs Sage 200 API access and **Company Access** to the company you want Hive to read. Ask your Sage administrator to grant both, then connect again. ### Installing Microsoft Teams fails The person installing the Teams channel must hold one of these Microsoft Entra roles: Global Administrator, Privileged Role Administrator, Cloud Application Administrator or Application Administrator. The Teams Administrator role alone is not enough. After consenting, the Hive app must also be added from inside Teams. See [Slack and Teams](/integrations/slack-and-teams). ### A custom MCP server's tools look out of date Resync the server from its card. Hive re-discovers the tools and pins their definitions again, which it does deliberately so a server cannot quietly change a tool after you connected it. See [MCP servers](/integrations/mcp-servers). ## Signals ### The Signals feed is empty An empty feed is **not an all-clear**: Hive can only raise signals from what it observes. 1. Open **Signals** and switch to **Coverage**. Each metric shows **Watching** (actively checking), **Learning** (building a baseline — a trend needs at least five readings) or **Paused** (turned off by your team). 2. Confirm the tool you expect is connected under **Integrations**. 3. Ask an Admin to select **Run loop now** rather than waiting for the next hourly run. 4. Check the source system actually contains something that should fire — Hive needs a sustained change, not a single spike. See [Coverage](/signals/coverage) and [Trustworthy signals](/signals/trustworthy-signals). ### A signal type keeps coming back after I dismiss it Dismissing with a reason teaches Hive. **Already handled** quiets that record, **This is expected — mute this type** stops the type for the workspace (Admins and Owners), and **Just dismiss it** changes nothing about future detection. See [Triage signals](/signals/triage-signals). ### I cannot act on a signal or approve an action Acting, snoozing, dismissing and approving need the **Operator** role or higher; **Analyst** is read-only. Each role also has a ceiling on how risky an approval it can make — Operator up to B2, Department lead up to B3, Owner and Admin up to B4. See [Roles and permissions](/admin/roles-and-permissions). ## Ask Hive ### The answer ends with "Live sources were not checked" Hive tried to read a connected tool live and the lookup failed, so the answer is based on stored Brain data that may be out of date. Check the connector's status, then ask again. Pinning the tool with `@` makes Hive read it directly. ### A model is missing or unavailable The picker only shows models your workspace can run. An Admin sees unavailable models with a reason — the provider is not configured, or the model lacks a capability the task needs — and can change the allowlist under **Settings → Models**. See [Models](/ask/models). ### The answer is weak or vague * Pin the right connector with `@`. * Name the customer, date range or metric. * Attach the document if the information is not in a connected tool. * Ask Hive to cite the source for each claim, and do not act on an answer that cannot. ## Workflows ### Deploy is refused Deploying checks the whole workflow first. Each problem comes back with a code; fix it in the builder and deploy again. | Code | What it means | Fix | | - | - | - | | `graph_invalid` | The graph breaks a basic rule: not exactly one trigger, no steps, a cycle, or an unreachable step. | Connect every step to the trigger and remove loops back to earlier steps. | | `condition_prompt_missing` | An AI condition has a blank condition. | Write the condition or delete it. | | `loop_unbounded` / `wait_unbounded` | A loop or wait has no usable limit. | Set **Max iterations** (up to 500) or a timeout (up to 30 days). | | `run_step_cap_exceeded` | The workflow could exceed 500 recorded steps in one run. | Lower loop iterations or retry counts inside loops. | | `expression_reference` | A `{{placeholder}}` points at a step that does not run before it. | Reference an earlier step's output. | | `branch_escapes_loop` | A branch leaves a loop's body. | Keep branches inside the loop, or move the step after it. | | `loop_parallel_unsupported` | A parallel loop contains a Wait, a Parallel agents step or another parallel loop. | Turn off **Parallel execution** or move that step out. | | `error_policy_reference` | A later step uses the output of a step set to **Continue with empty output**. | Change the error setting or stop referencing that output. | | `error_policy_unsupported` | A retry or continue setting is on a step that cannot use it. | Use the default error setting on that step. | | `output_key_conflict` / `output_key_unknown` | Two Output steps share a key, or a key is invalid. | Give each Output a unique key. | | `web_search_query_placeholder` | A fixed web search query contains `{{…}}`. | Switch the step to **Automatic**, or remove the placeholder. | | `action_forbidden_role` | An action is above your workspace role. | Ask someone with the right role to deploy, or remove the action. | | `action_not_authorable` | The action type cannot be used in a workflow. | Use a different step. | | `agent_tool_unavailable`, `web_search_unavailable`, `image_generation_unavailable`, `document_files_unavailable`, `saved_agent_unavailable`, `workflow_model_unavailable` | A capability the step needs is not available in your workspace. | Pick another model or step, or ask your Hive account team. | | `saved_agent_public_form_unavailable` | A saved agent step is in a workflow started by a public form. | Use a Hive agent step, or a different trigger. | | `connector_capability_missing` | A connector step needs an app you have not connected. | Connect the app under **Integrations**. | See [Test and deploy](/workflows/test-and-deploy) and [Runs and errors](/workflows/runs-and-errors). ### "This form is not available" A public form link stops working when a new link is created, the workflow is redeployed, re-activated, paused or deleted, or its owner loses the right to run it. Open the workflow, select **Form link**, create a new link and share that. See [Schedules and forms](/workflows/schedules-and-forms). ### A schedule stopped running A schedule runs as the member who created it. If a different member's release becomes live, the schedule waits for its owner. Check **Scheduled** for its status and next run. ### Runs are paused for the rest of the month Your plan's Execution allowance is used up: new agent and workflow runs pause until it resets. An Owner or Admin can move to a larger tier under **Billing**. See [Billing and plans](/admin/billing-and-plans). ## Billing and access ### "Your trial has ended" The 14-day trial is over. Your data is safe. An Owner or Admin can subscribe from the paywall or **Settings → Billing** to restore access; other members are asked to contact an admin. See [Trial and payments](/admin/trial-and-payments). ### "Payment past due" or "Billing actions are restricted" A payment failed. Hive retries the payment method on file, and after a seven-day grace period billable actions are restricted. An Owner or Admin can update the card from **Billing**; access returns once payment is recovered. ### I cannot see a screen Some settings are for Owners and Admins only, some products can be switched off for the workspace, and your role may be limited to certain departments. Ask an Owner or Admin to check your role in **Settings → Members**. ## Actions are halted A banner reading "**Kill switch is on**" means every autonomous action and approval is stopped company-wide. Owners and Admins can select **Resume autonomy** — but only once the reason it was engaged is understood. See [Safety and control](/admin/safety-and-control). ## Before you contact support Collect: * the screen, the exact message and the time it happened; * your role; * the connector, workflow, signal or conversation involved, with its link; * the Audit receipt for an action problem; * whether **Retry** or **Run loop now** changed anything. Never include passwords, tokens, webhook secrets or customer data. Then follow [Get support](/help/support).