Invite-only. Ask your administrator for credentials, then you are one command away. Quickstart →

The terminal assistant for developers

Straight answers for developers, right in the terminal.

Ask a question or pipe in broken code. T Chat replies with the fix, the command or the snippet and nothing else. One install, a username and a PIN.

Install Read the docs

Works inWindowsUbuntumacOSWSLGit BashSSH

Ask from anywhere

One command to install. Then ask from any shell, on any machine.

No admin rights

Installs into your home folder and your user PATH. Works on locked-down lab machines.

Pipe anything in

Send a file, a diff or an error with cat, type or a pipe, and add the question as an argument.

Chat when you need it

Run ask with no arguments for a conversation that remembers the thread.

terminal

$ cat app.py | ask "fix the bug"
for i in range(10):
    print(i)
Cause: missing colon after range(10).
 
$ ask "undo last git commit, keep changes"
git reset --soft HEAD~1
 
$

Fix it

Code in, fix out. Pipe a file, a diff or an error and get the corrected code with a one-line cause.

Straight answers

No greetings, no preamble, no recap of your question. The fix, then the reason.

Clean code

Generated code comes without comments unless you ask for them.

Exact commands

Ask how to do something in a shell and get the command, not an essay.

app.py

$ ask "why does this return None?"

The else branch has no return. Add return result.

server.js

$ ask "fix the unhandled rejection"

Wrap the await in try/catch and send a 500 from catch.

query.sql

$ ask "speed this up"

Add an index on (user_id, created_at); drop SELECT *.

Dockerfile

$ ask "make the image smaller"

Use a multi-stage build and a slim runtime image.

deploy.sh

$ ask "why does this fail on Ubuntu?"

CRLF line endings. Run dos2unix deploy.sh.

Run it for a team

Every member gets a PIN. Limits, usage and access in one console.

Invite-only

You create each account. Members sign in with a username and a 4-character PIN, and can change it themselves.

Limits per user

Set a daily cap and a per-minute rate for each person. Over the limit, no model call is made.

Usage you can see

Questions per user and per day, last activity, and one-click disable or PIN reset.

T CHAT / Console

24h15

7 days115

Active2 / 3

rrahulActive8 / 50

ppriyaActive6 / 30

aarjunDisabled1 / 50

How a question travels

Checked first, then answered. Every question passes the same gates before a model sees it.

Authenticated

Each request carries a revocable token. Disable a user or change a PIN and it stops working at once.

Rate limited

Per-minute and rolling 24-hour limits are checked before any model call.

Streamed

The reply arrives token by token, plain text, ready to copy.

Terminal

ask "why does this return None?"

Checks passed ›

The else branch has no return statement. Add return result.

ask usage

Checks passed ›

last 24h: 12 / 50
rate: 5 per minute

Built to be trusted

PINs are locked down

Hashed with scrypt. Five wrong tries lock a username for 15 minutes, and repeated failures lock the source address.

Tokens, not passwords

The terminal stores a random token. Only its hash is kept on the server, and it can be revoked.

Usage, not prompts

T Chat stores counts and sizes, never your questions or answers. The model provider still receives them.

Your model, your key

Runs on any OpenAI-compatible API. The provider key lives only in server environment variables.