Warp

warp.devcontributed by Samuele Ongaro

ALMOST

A weekend of work, and real gaps remain.

A terminal rewritten around blocks: each command and its output is a unit you can copy, share or search, with completions and a command palette.

Promptfree, for everyone, and the only version there is
Build me a terminal that replaces Warp — and be honest about why this is ALMOST.

**A rendering engine built for the terminal** is what makes it feel different, and it is a real piece of engineering: a GPU-accelerated grid that stays at sixty frames a second while a build floods it with output. Existing open terminals already solved that — use one, and build the layer above it. That is the honest route, and it is a fortnight rather than a year.

STACK
- Start from an existing GPU-accelerated terminal, or from a mature terminal library. Do not write a renderer or a VT parser
- The layer you add: a shell integration, a block model, and a command history that is worth searching
- SQLite on device for history
- No server, no account, no network. A terminal sees every command and every secret you paste

THE BLOCK MODEL, WHICH IS THE IDEA WORTH COPYING
- A command and its output are one thing: a block, with a start, an end, an exit code and a duration
- That needs a shell integration emitting a marker before the prompt, before the command runs, and after it finishes. The escape sequences for this are standard and every modern shell supports the hooks
- With blocks you get: click to select a whole command's output, copy just the output, scroll by command rather than by line, collapse a noisy build, and jump to the last failure
- Jumping to the previous failed command is the single most useful thing a terminal can do and almost none of them can

HISTORY WORTH HAVING
- Every command stored with its directory, its exit code, its duration, the host, and the shell session
- Search across all of it, instantly, filtered by directory or by exit code. 'The ffmpeg command I ran in this folder last month that worked' is then a search rather than an archaeology exercise
- Ranking by frecency and by directory match, so the suggestion in this project is a command you ran in this project
- **Never store a command containing something that looks like a secret.** Match on patterns and on well-known flags, and offer a per-session pause. A history database is a credential store nobody intended to build

INPUT THAT BEHAVES
- Multi-line editing that works like a text editor: arrow keys, selection, word movement, undo. A long pipeline is easier to write with a cursor that goes where you point it
- Completion from the history, from the filesystem, and from a shell's own completion where it can be reached
- Never rewrite what the user typed. Suggest inline, accept explicitly, and let the shell see exactly the bytes that were entered

WHAT TO GET RIGHT BECAUSE PEOPLE NOTICE
- Latency from keystroke to glyph. This is felt constantly and is the reason a terminal is loved or abandoned. Measure it and treat a regression as the highest-severity bug
- Correct handling of everything the terminal is asked to do: 256 colours and true colour, mouse reporting, bracketed paste, alternate screen, wide and combining characters, right-to-left text, and reflow on resize. These are the details that break full-screen applications, and full-screen applications are what serious terminal users live in
- A ligature-capable font with proper fallback for symbols and non-Latin text
- Bracketed paste on, always, and a warning before pasting anything with a newline in it — that one has caused real accidents

SESSIONS
- Split panes and tabs, with a saved layout per project directory
- Reconnect to a long-running remote session; do not reimplement one — the existing multiplexers are better than anything you will write
- A command palette, because that is how anything not bound to a key gets found

WHAT NOT TO BUILD
- A cloud account, shared sessions, or anything that sends a command to a server. The moment a terminal has a network component it becomes a thing to audit
- If you want a command explained by a model, run it locally and say clearly what is sent where

OPERATIONS
- Configuration in a plain file
- History exportable and deletable in one command
- No telemetry. Not one event

WHAT MATTERS MOST
Blocks and latency. The shell integration is a day and it changes how the terminal feels; the rendering you should inherit rather than write.

What you lose

  • A GPU-accelerated renderer with block-structured output, which is a serious piece of engineering
  • Completions and command suggestions across many tools
  • Shared blocks and team workflows on the paid tier

If you would rather not build

  • tmux with a shell history hook
  • Your existing terminal plus fzf

What it costs

read from their page 15 Aug 2026

PlanBilled monthlyBilled yearlyLast read
—$20/mo—15 Aug 2026

Their pricing page is where these came from. Seeing a different price? Tell us.

The escape hatch

open source · no votes, no paid placement

WezTerm

$0

A GPU-accelerated terminal with multiplexing built in.

wezterm/weztermfree · open source

Ghostty

$0

A fast native terminal focused on correctness and speed.

ghostty-org/ghosttyfree · open source

Why this verdict

our own opinion · changed only by a person

52/100

Verdict kinda at 52. Free terminals and a shell hook cover most of the daily value; the block interface itself is not a weekend build.

History

tracked since 10 Aug 2026 · nothing is ever overwritten

Interest · last 30 dayspeak 1/day
views0130 Aug4 Sept9 Sept14 Sept19 Sept24 Sept28 Sept
— views— prompt copies none yet— votes none yet

Questions about Warp

answered from the record above

Is Warp free?

No — the plan we track is $20 a month. Build from $20/month, pay as you go on AI usage; a free tier covers the terminal itself with a small AI allowance.

Can you replace Warp by building your own?

ALMOST. A weekend of work, and real gaps remain. Replacement score 52 out of 100, build time a weekend. Read what you lose before you decide.

How much does Warp cost?

$20 a month on Pro — $240 a year. Recorded 14 Aug 2026.

What do you lose by replacing Warp?

A GPU-accelerated renderer with block-structured output, which is a serious piece of engineering; Completions and command suggestions across many tools; Shared blocks and team workflows on the paid tier. If any of those carry weight for you, keep paying.

Is there an open-source alternative to Warp?

Yes: WezTerm, Ghostty. The prompt on this page is for when you want it your way instead.

Related entries

same category first, most replaced first

All 56 in Dev tools

Not sending yet

Every week, something stops being worth paying for.

New verdicts, prices that moved, entries added. One email a week. Unsubscribe in one click. Nothing is being sent yet — your address is kept here, and the first issue is the first thing it is used for.

free forever · no tracking pixel · stored here, never passed to anyone

Esc