TRMNLY
01 / 12Book a demo

TRMNLY · Sales deck · October 2026

The terminal your agentsalready live in.

← → to move · F for fullscreen

~/code/shop
~/code/shop main ✓
$ npm test
PASS src/cart.test.ts
PASS src/search.test.ts
FAIL src/checkout.test.ts
● applies a discount code before tax
Expected: 90.00
Received: 99.00
Tests: 1 failed, 23 passed
~/code/shop main ✱
$
SON~/code/shopClaude Sonnet 5.5
The checkout test is failing. Fix it.
Read terminalpane 1done
Edit filesrc/checkout.tsdone

The discount was applied after tax. I moved it ahead of applyTax(); rerunning the checkout suite to confirm.

Run in panenpm test -- checkoutapproval needed
$ npm test -- checkout
in pane 1 · ~/code/shop
ApproveDenyEdit argsAlways… ⌄
Message the agent…
The problem02

Developers use agents. They do not trust them yet.

84%

of developers use or plan to use AI tools, up from 76% a year earlier[1]

81%

of developers using AI agents have security or privacy concerns about them[1]

6.6%

trust AI for important decisions; about half trust it only when they can verify the result[2]

The product03

A real terminal first. Everything else folds away.

~/code/shop
~/code/shop main ✓
$ npm test
PASS src/cart.test.ts
PASS src/search.test.ts
FAIL src/checkout.test.ts
● applies a discount code before tax
Expected: 90.00
Received: 99.00
Tests: 1 failed, 23 passed
~/code/shop main ✱
$
SON~/code/shopClaude Sonnet 5.5
The checkout test is failing. Fix it.
Read terminalpane 1done
Edit filesrc/checkout.tsdone

The discount was applied after tax. I moved it ahead of applyTax(); rerunning the checkout suite to confirm.

Run in panenpm test -- checkoutapproval needed
$ npm test -- checkout
in pane 1 · ~/code/shop
ApproveDenyEdit argsAlways… ⌄
Message the agent…
Reach04

400+ tools across the whole machine.

Shell & panes

Read any pane, run commands you can watch, queue work across splits and tabs.

SSH, Mosh & SFTP

Your saved hosts, port-forward presets and file transfers.

Containers

Docker and a built-in microVM engine, with shells into either.

Ports & processes

Find what is listening, reserve ports per project, kill a stuck listener.

Browser

An in-app browser the agent can drive, with its console and network log.

Git & worktrees

Parallel worktrees per branch, diffs, review comments and PRs.

The window05

One window. The whole machine.

1 · Left sidebar

Projects, and every tab with the agent it is running and its status.

2 · Terminal

Ghostty, unchanged. The agent reads the pane that failed and runs in it.

3 · Right sidebar

Agent, prompts, queue, files and browser; each folds away when you don't need it.

Keep your agent06

Already on Claude Code? Keep it.

Claude Code Codex Gemini CLI Ollama OpenRouter

Most-loved AI tool among engineers[3]

Claude Code46%
Cursor19%
Copilot9%
Safety07

Approvals you can actually tune.

Posture per project

Ask, plan or auto-accept, switched with one key and kept across restarts.

Read and write, labelled

Every tool is tagged, so approvals can be broad for reads and narrow for writes.

Rules scoped to hosts and paths

Allow a command on staging and still be asked on production.

A gate for destructive verbs

It fires even on auto-accept. So does any tool you mark always-ask.

Audit log and write snapshots

See what ran and revert a file the agent changed.

Agent Lock

Blocks keyboard and mouse input and covers the screens while an agent works.

Approval · pane 2
Run on prod-1?
sudo systemctl restart api
Allow onceAlways on stagingDeny
allowgit status · npm test
askgit push · ssh prod-*
denysudo · rm -rf ~
Anywhere08

Your Mac from your phone. A Linux box when you need one.

Approval
git push on your Mac?
AllowDeny

iPhone remote

Pair an iPhone with your Mac and drive panes and agents over an encrypted tunnel. Approve a risky command from your pocket. Public beta on TestFlight.

$ claude
● 2 vCPU · 4 GB
files kept
0.5 cr / min

Cloud workspace

A real Linux machine with a real shell, files kept between sessions, and the agent toolchain installed. It bills by the minute only while a terminal is open on it.[8]

Pricing09

No markup on a single token.

$99

per seat for the desktop app, a year of updates included — keep it forever, renew for $69 a year. Model tokens go from your Mac to the provider at their rates.[8]

$0.50

an hour for a cloud workspace, falling to $0.40 on larger packs[8]

100

free cloud credits on a new account, no card[8]

See full pricing →

How it compares10

Where TRMNLY sits.

TRMNLYWarpConductorCursor
What it isNative terminal + agentAgentic terminalMac agent orchestratorAI code editor
Terminal coreGhosttyRebuilt from scratchNoneEditor-embedded
Agent reachWhole machineTerminal and repoRepo worktreesRepo
ModelsYour keys or CLIs, no markupCredits; own key optionalYour CLIsUsage at API rates
Entry price$99 a seat, renew $69/yr; cloud $0.50/hr[8]$20/mo Build[4]Free; Pro $50[5]$20/mo Pro[6]

Competitor descriptions are our reading of their public pages on the dates listed under Sources. Product names and logos belong to their owners; no endorsement is implied.

Next step11

See it on your own machine.

TRMNLY is available today on macOS, with early-access accounts approved as they come in. A demo starts with an infrastructure problem of yours, not a toy repo.

Sources12

Sources.

  1. [1]2025 Developer Survey: AIStack Overflow · July 2025
  2. [2]2026 Developer Survey press release: AI trust is conditionalStack Overflow · 6 October 2026
  3. [3]AI Tooling for Software Engineers in 2026The Pragmatic Engineer · 2026 (survey run 27 Jan–17 Feb 2026)
  4. [4]Warp pricingWarp · checked 27 September 2026
  5. [5]Conductor pricingConductor · checked 27 September 2026
  6. [6]Cursor pricingCursor · checked 27 September 2026
  7. [7]Ghosttyghostty.org · 2026
  8. [8]TRMNLY pricingtrmnly.dev · October 2026

Figures as of 19 October 2026.