# Troubleshooting > **A chapter to look things up in.** Find your symptom. ## First — how to read the log Most problems have their answer in the log. ```bash cat logs/server.log ``` Just the end: ```bash tail -50 logs/server.log ``` If it died on the previous run: ```bash cat logs/server.log.prev ``` --- ## Install and startup ### `hyperteams: command not found` **The most common problem.** 1. **Open a new terminal window.** The install changes your path settings and an already-open terminal does not know. 2. If that fails, restart the computer 3. If that fails, the install did not complete → [install again](/guide/ht-install) ### It starts and immediately stops Check the log. Usually a prerequisite problem. ```bash claude --version # does it print? node --version # does it print? ``` If either does not, fix [prerequisites](/guide/ht-prerequisites) first. ### The port is already in use ```bash PORT=9000 hyperteams ``` To change the default, `hyperteams setup`. ### The address prints but the browser will not open it - Did you type it exactly, including `http://` - Is a firewall blocking it - Try another port → `PORT=9000 hyperteams` --- ## Access and authentication ### I do not remember the password ```bash hyperteams setup ``` ### I connected the tunnel but it will not open from outside 1. **Did you restart?** `hyperteams restart` is required after connecting a tunnel 2. Did you use the exact address 3. Does the dashboard settings screen show the tunnel as connected --- ## Running tasks ### I handed over a task and nothing happens ``` □ Is the Claude Code login still valid? → run claude in a terminal □ Is the working directory path valid? → does the folder actually exist □ Are there errors in the log? → tail -50 logs/server.log ``` **The first is the most common.** When the Claude Code login expires, HyperTeams stops with it. ### It touches the wrong files Your [working directory is probably drawn too wide](/guide/ht-working-directory). Redraw it narrowly. And **state the scope** in your instruction — [writing it as what not to do is effective](/guide/ai-intent-context). ### I keep disliking the results The cause differs by symptom. | Symptom | Cause | Response | |---|---|---| | Only generic answers | missing context | add the situation to the folder description and the instruction | | It forgets earlier instructions | [context pushed out](/guide/ai-context-overflow) | split into a new task | | It does what you said not to | scope unspecified | state what not to do | | Unclear whether it finished | no done condition | add a pass criterion | ### The job runs far too long The task is too big. [Split it smaller](/guide/ha-micro-sprint) — one task fitting in a single sentence is about the right size. --- ## Performance and resources ### The machine gets slow ``` □ have you left several terminals open and forgotten them □ are too many tasks running at once ``` Clean up terminals you are not using. ### Costs are higher than expected ``` □ how many scheduled runs are going □ are there schedules left over whose purpose is gone □ are you using a heavy model for simple work ``` See the cost section of the [scheduled runs chapter](/guide/ht-schedule). --- ## Data and recovery ### My records disappeared `hyperteams upgrade` does not touch data. If you reinstalled, data may have been removed. ### The work went wrong and I want to undo it **If you use git**, undoing there is the most reliable route. That is why [committing before starting](/guide/ht-autonomy) is recommended. Without git you depend on your editor's undo or a backup. --- ## If none of that works Get in touch, with the log. ```bash tail -100 logs/server.log ``` Contact: **hyperteamsnet@gmail.com** --- Walk the recovery order once, for yourself. Thirty minutes → [Recover a stalled job](/guide/ht-try-recovery) Including these speeds things up: ``` □ OS and version □ hyperteams --version □ claude --version / node --version □ what you were trying to do and what did not work □ the end of the log ```