Build agentic workflows at ease.

Flows, Celi, Atlas, Kanban, Review and Code — one local-first platform where agents build, review and record your software.

Download
or install from your terminal
curl -fsSL https://www.celial.ai/install.sh | sh

Local-first · Your keys, your models · Free to start

Meet Celial

Everything your agents do. All on one canvas.

Compose, run, and review whole fleets of agents without losing the thread.

Real captures — Celial, Tahel appearance

The shift

Development, before and after agents.

One PR at a time

Ten in flight

Babysitting a terminal

Reading a review

Rerun from zero

Resume from the failed node

Guessing what's left

Tokens and the window, per provider

Proof

Agents were a black box. Now every app shows its work.

Celi's plan canvas — the steps of a proposed workflow drawn as nodes beside the conversation, with an Approve plan button above them

Plans you can see

Celi, the assistant inside every app, lays its plan on a canvas before it acts, asks for what it needs through one gate, and remembers across conversations.

The Kanban board — To do and In Progress columns, each ticket still carrying the screen snap it was made from

Ctrl+B: snap it, it becomes a ticket

Snap anything on your screen, say what's wrong, and a triage agent writes the ticket, files it on the board, and hands it to a workflow.

The Atlas Library — features grouped by area, each row a real capability with the sentence that describes it

The company's record

What the product does, where it lives in your checkout, what was decided and when it shipped — kept current by the agents doing the work.

Token and limit usage — one row per provider: how much of the five-hour and weekly window is spent, when it resets, and the tokens each one used this week

Tokens, and whether you can keep working

Every provider window in one place, under your profile: what's left, what used it, and which run spent it.

Local-first

Runs on your machine. Answers to no one else.

Five agentic CLIs

Claude Code, Codex, Antigravity, Grok and OpenCode — one seam over all five, on your own subscriptions and keys, with the model chosen per node.

Your work stays put

Agents execute locally, in terminal sessions Celial owns on your machine. Sync is optional, and what it carries is counts — never your prompts or your code.

Versionable workflows

Readable, diffable workflows — version them like the code they write.

Control

Powerful runs, kept on a leash.

Gates & questions

Put a gate on any agent node — the run pauses for your approval, and agents can ask you questions while they work.

Resume from failure

A failed node doesn't restart the run. Fix it and resume exactly where it stopped.

Tokens while it happens

Token use rolls up per node and per run, against each provider's window — visible during the run, not after.

Early words

Built with the people who run agents all day.

“I stopped babysitting terminals. I design the run, gate the risky nodes, and read the review.”

Early access user

Staff engineer

“The first agent tool where I always know what's running — and how much of the window is left.”

Early access user

Platform lead

“Ten branches in flight before lunch. Review is the job now, and that's the point.”

Early access user

Senior developer

“The canvas is the mental model I always had in my head. Now it's the actual interface.”

Early access user

Tech lead

“Gates mean I trust it with real work. Nothing ships without me.”

Early access user

Engineering manager

“It made running agents in parallel feel boring — in the best possible way.”

Early access user

Indie developer

Develop at the speed of agents.

Download Celial and put your agents to work — local-first, on your machine, in minutes.

Download

Start now for free — no card. macOS, Linux, Windows.

The Celial hub — a card for each Celial app, with Celi's agents in the dock below.