# Errors > The error messages DevHub tools return and what to change in the next call. A DevHub error says what to fix. Fix the argument and call again. One error is never a reason to stop using DevHub for the rest of the session. ## Access and availability | Message | What it means | What to do | | ----------------------------------------------------------- | ------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------ | | `sign in required` (on every request) | the user's session ended; the server stopped serving | tell the user to sign in to DevHub again. Do not retry in a loop | | `MCP tool "" is disabled in Settings` | the user switched the tool off | use another tool or ask the user to enable it in Settings > MCP Server > Allowed MCP tools | | `MCP tool "" is not available: requires ` | the user's Vartari plan does not include this area for agents, or not at the level the tool needs | do not retry. Tell the user which permission it names | | `MCP tool "" is not enabled for this endpoint` | this endpoint's profile does not offer the tool | use a tool you were offered | | `Unknown tool` | the name does not exist | take names from the tool list, not from memory | | ` is not configured` or `is not available` | the feature is not set up in DevHub | tell the user; nothing to fix in the call | ## Arguments | Message | What to do | | ------------------------------------------------------------------------- | --------------------------------------------------------------------------------------- | | `A claim needs an owner` | pass `by` to `plan_claim` or `plan_renew` | | `query is required` | pass a non-empty `query` | | `workspaceId is required when saving a screenshot` | pass `workspaceId` with `save` | | `wait must be seconds from 0 to 45` | pass `wait` between 0 and 45 | | `allowAgents must be true or false` | pass a boolean | | `Request too large` | send less in one call; split the content | | a hash or text mismatch on `file_edit`, `file_replace`, `vault_file_edit` | read the file again and rebuild the edit from the new text | | `already exists` on a create | the file is there. Read it; use `file_replace` or `file_write` if replacing is intended | | a script is refused | `npm_script_list` shows what is declared. Only declared scripts run | ## Vaults | Message | What to do | | ------------------------------------------------ | ---------------------------------------------------------------------------------------------------------------------------- | | `Vault not found or disabled` | take `vaultId` from `vault_list` | | `Vault is not configured for MCP write access` | the vault is read-only for MCP. Ask the user to allow writing, or write elsewhere | | `This note is in a knowledge folder: ` | the note fails the rules of that folder. Fix what the message lists and write again | | `knowledge_search` returns nothing | a miss is not absence. `vault_overview`, then `vault_files_search`; `knowledge_index_status`, then `knowledge_index_refresh` | ## Agent endpoints An agent started by DevHub works under narrower rules. | Message | What to do | | --------------------------------------------------------------------------------------- | -------------------------------------------- | | `This agent works only in its own repository. Use its workspaceId.` | use the `workspaceId` of your own repository | | `This agent may use only its own repository's, project's and organization's vaults.` | take a `vaultId` from `vault_list` | | `This plan is not for this agent's repository.` | take plans from `plan_ready` | | `An agent cannot start itself. Start a different agent, or answer the task yourself.` | answer the task, or start another agent | | `An agent cannot give another agent leave to start agents. Run it without allowAgents.` | leave `allowAgents` out | | `An agent cannot close an investigation. …` | say in your answer that it is ready to close | ## Agents and runs | Message | What to do | | ------------------------------------------------------------------------- | -------------------------------------------------------------------- | | `No such agent on this endpoint. Call agent_list for the ones there are.` | take `agentId` from `agent_list` | | `No persona ""` or `No client ""` | `agent_list` shows the ones `agent_create` accepts | | `No such run` | take `runId` from the `RUNNING` answer. A run's answer lasts an hour | | `RUNNING ` | not an error. Call `agent_run_status` with `wait` | ## Plans | Message | What to do | | --------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- | | a refused `plan_claim` | the refusal says why: someone holds the step, a gate is open, or a step in progress touches the same files. Call `plan_ready` and take another step | | `Step not closed: ` | the working tree did not merge. Resolve the conflict in the working tree, commit, call `plan_close` again | | a refused `plan_close` | you are not the holder, or the step is not in a state that can close. Read the message; `step_resume` shows the state | | refused `plan_create` | the id is taken. Choose another or omit it | ## Finding the cause of a failure - `audit_list` with `tool` and `result` shows the failed call: arguments, error, timing. - `audit_analyze` shows repeated errors and slow tools. - `audit_summary` shows the success rate over a window. ## Related - [Limits](/devhub/for-agents/limits/index.md) - [Troubleshooting](/devhub/troubleshooting/index.md)