# Messages you may see > Every fixed message LLMGRAPH can show when it cannot answer, read from the program's own message tables, with what you can do. When LLMGRAPH cannot answer, it replies with a fixed message instead. Every message below is read from the program's own message tables when this page is built, so the words are the ones your agent sees. Words in braces, such as {code}, are filled in when the message is shown. Many messages tell the agent to ask you. Each entry says what you can do. ## Sign-in, licence and plan messages These begin with N. The command shown in them is the launcher string the program holds in one place. ### N1 > llmgraph · not signed in on this computer. Ask the user to open {link} and approve only if they just started llmgraph on this computer; the code {code} must match the page. Do not retry until the user confirms. What to do: Open the link and check that the code on the page matches the code in the message. Approve only if you just started LLMGRAPH on this computer. ### N1b > llmgraph · not signed in, and the licence server cannot be reached. Ask the user to run "npx -y llmgraph-mcp login" when online. Do not retry. What to do: Connect to the internet, then run the login command shown in the message. ### N1c > llmgraph · off: this computer's sign-in is no longer accepted by the licence server. Ask the user to run "npx -y llmgraph-mcp login". Do not retry until the user confirms. What to do: Run the login command shown in the message to sign this computer in again. ### N1d > llmgraph · not signed in on this computer. Ask the user to run "npx -y llmgraph-mcp login". Do not retry until the user confirms. What to do: Run the login command shown in the message, then approve this computer in your browser. The message does not mean the licence server is down; it says only that this computer is not signed in. ### N2 > llmgraph · waiting for the user to confirm the sign-in (code {code}). Do not retry until they do. What to do: Finish the sign-in in your browser; the agent waits until you confirm. ### N3 > llmgraph · off: the licence ran out and the licence server cannot be reached. Reading and editing your code is unaffected. Retry once the connection is back; do not retry in a loop. What to do: Reading and editing your code is unaffected. Reconnect to the internet; the licence renews on its own. ### N4 > llmgraph · off: this account has no active subscription or trial. Ask the user to subscribe or renew at {account_url}. What to do: Open the account page and start or renew a subscription or trial. ### N4b > llmgraph · off: the free trial for this account was already used. Ask the user to subscribe at {account_url}. What to do: The trial can be used once per account. Subscribe on the account page. ### N5 > llmgraph · off: this computer is not licensed on the account (it was removed, or the account is paused). Ask the user to check {account_url}; if the account is fine, "npx -y llmgraph-mcp login" signs this computer in again. Do not retry. What to do: Check the devices list on the account page. If the account is fine, run the login command in the message. ### N6 > llmgraph · off: the account is at its device limit{n_part}. Ask the user to remove a device at {account_url}. What to do: Remove a device on the account page, then try again. ### N7 > llmgraph · off: version {v} is too old (minimum {min}). Ask the user to run "npx -y llmgraph-mcp@latest". Do not retry. What to do: Run the command in the message to update, then restart your agent. ### N8 > llmgraph · off: this computer's clock is about {minutes} minutes {direction} the licence server. Ask the user to correct the clock, then retry. What to do: Set your computer's clock correctly (automatic time is best), then try again. ### N9 > llmgraph · not indexed: this repository has {n} code lines and the plan allows {max}. Other repositories are unaffected. What to do: This repository is over the line limit of your plan. Other repositories are not affected. ### N10 > llmgraph · not indexed: the plan allows {max} repositories on this computer and {n} are already indexed. Ask the user to remove one with "npx -y llmgraph-mcp repos", or change plan at {account_url}. What to do: Remove one indexed repository with the command in the message, or change plan on the account page. ### N11 > llmgraph · note: the saved index could not be opened on this computer, so it was rebuilt. This first answer is slower. What to do: Nothing to do. The index was rebuilt, so this first answer is slower. ### N12 > llmgraph · off: the system keychain is not available, so the index cannot be protected. Ask the user to run "npx -y llmgraph-mcp status". What to do: Make the system keychain available, then run the status command in the message. ### N13 > llmgraph · off: the installation looks damaged or changed. Ask the user to reinstall with "npx -y llmgraph-mcp@latest". What to do: Reinstall with the command in the message. ### N14 > llmgraph · note: the licence server cannot be reached; answers stop in about {n} minutes unless the connection returns. What to do: Nothing to do if the connection returns in time. Answers stop when the current licence runs out. ### N15 > llmgraph · note: your subscription ends in {n} day(s). Ask the user to renew at {account_url}. What to do: Renew on the account page before the subscription ends. ### N16 > llmgraph · off: the licence could not be verified. Ask the user to run "npx -y llmgraph-mcp status"; if it persists, reinstall. What to do: Run the status command in the message. If it persists, reinstall. ### N17 > llmgraph · off: a newer LLMGRAPH saved the index of this project. Ask the user to update with "npx -y llmgraph-mcp@latest". What to do: Update LLMGRAPH with the command in the message. A newer version saved this project's index, and an older one never overwrites it. ## Messages about a call or the project These begin with M. ### M1 > llmgraph · bad input: "{f}" is not a field of {tool} (fields: {list}). What to do: The agent sent a field that question does not have. The agent corrects the call; nothing for you to do. ### M2 > llmgraph · bad input: "{f}" is required. What to do: The agent left out a required field. The agent corrects the call. ### M3 > llmgraph · bad input: "{f}" must be {type}. What to do: The agent sent a value of the wrong type. The agent corrects the call. ### M4 > llmgraph · bad input: "{f}" is empty. What to do: The agent sent an empty value. The agent corrects the call. ### M5 > llmgraph · bad input: "{f}" is too long (limit {n}). What to do: The agent sent a value that is too long. The agent shortens it. ### M5b > llmgraph · bad input: "{f}" has more than {n} entries. What to do: The agent sent too many list entries. The agent sends fewer. ### M6 > llmgraph · bad input: "{p}" is outside this project. What to do: The path the agent named is outside the project. Ask the agent for a path inside it. ### M7 > llmgraph · {h} is not from this session (handles reset when the server restarts). What to do: The handle is from an earlier session. Ask the question again to get a new handle. ### M8 > llmgraph · cannot tell which project to read: this server started in a folder that is not a project (the user's home folder or a drive root) and the agent client named none. Ask the user to add "--repo " to the server command. What to do: Add the repository option to the server command in your agent's settings, naming your project folder. ### M9 > llmgraph · the project folder named in the server settings ("--repo" or LLMGRAPH_REPO) cannot be used: it does not exist, is not a folder, is the user's home folder or is a drive root. Ask the user to correct it. What to do: Correct the project folder named in the server settings (the repository option or the environment variable). ### M10 > llmgraph · the project folder the agent client reported cannot be used (it is not a folder on this computer, or it is the user's home folder or a drive root), and this server did not start in a project either. Ask the user to add "--repo " to the server command. What to do: Add the repository option to the server command, naming your project folder. ### M11 > llmgraph · still reading this repository ({k} of {n} files). No answer yet; ask again in a moment, not in a loop. What to do: Wait a moment and ask again once. Do not ask in a loop. ### M11b > llmgraph · starting the type checkers (the {n} files are read). No answer yet; ask again in a moment, not in a loop. What to do: Wait a moment and ask again once. Do not ask in a loop. ### M12 > llmgraph · stopped after {s} s: this took longer than the limit. Your files are untouched, and the index kept what it finished, so asking again may succeed. What to do: Ask again; the index kept what it finished, so a second try may succeed. ### M13 > llmgraph · internal error (ref {ref}). Your files are untouched. Try once more; if it repeats, tell the user. What to do: Try once more. If it repeats, tell us and include the reference shown in the message. ### M14 > llmgraph · "{name}" is not an enabled question (enabled: {list}). What to do: The agent asked a question that is not enabled. It should use one of the questions listed in the message. ### M15 > llmgraph · another llmgraph{who} is using the saved index of this project. Ask the user to close the other agent session, or run "npx -y llmgraph-mcp stop". The braces `{who}` are filled with `(PID , started UTC)`, or left empty when the other engine cannot be found. What to do: close the other agent session, or run `npx -y llmgraph-mcp ps` to see it and `npx -y llmgraph-mcp stop` to end it. The waiting session tries again every 2 seconds, so it recovers once the other one lets go. See [how it fits](https://beta.llmgraph.in/docs/how-it-fits.md). ### M16 > llmgraph · this project has no Python or TypeScript/JavaScript files that LLMGRAPH reads (files in other languages: {n}). Answers would be empty, so none are given. What to do: Nothing is wrong with the installation. LLMGRAPH reads Python, TypeScript and JavaScript files; open a project that has some, or add the folder that holds them. ### M17 > llmgraph · a newer LLMGRAPH is running on this computer. Ask the user to restart this agent session. What to do: Restart the agent session. LLMGRAPH was updated while this session was open, and the engine that is running now belongs to the newer version; a session of the older version never reads or changes the saved index of a newer one. ### M18 > llmgraph · the engine restarted; handles from earlier answers are reset. This line comes first in the next answer after the engine restarted. What to do: nothing. The engine that keeps your project read between agent sessions ended and started again, so the handles the earlier answers gave out (for example `w3`) are no longer known; ask the question again to get new ones.