Sign inContinue with Google

Cloud AI coding teams for your repository

Give your backlog
an AI team.

Lispa is a cloud workspace where teams of AI agents build, test, and review work in your repository. Powerful models take the hard parts, lower-cost models the routine steps. Start with one real task from your backlog.

Get startedSee how it works

$20/month includes $20 in usage credits.
Unlimited teammates · Choose a plan after signup · Cancel any time

A developer hands one task to Lispa. A coordinator returns a finished pull request with passing checks while Claude, Codex and Grok agents build, test and review on cloud machines.
You set the goal. They build, test, and review.Illustration · Provider marks show compatible agents · See the actual Lispa interface ↓

Bring the models
you already use.

  • Claude
  • Codex
  • Grok
  • Gemini
  • Open models
Your connected plans, API keys,
or Lispa credits.
The actual interface

This is what Lispa looks like.

Real screens from the Lispa console with demo data. Open any of them full size.

One place for the brief. The goal, the work in progress, and what needs your review.Enlarge screen
Every task has an owner. In progress, in review and done, with the agent or team on each.Enlarge screen
Evidence next to the change. The agent review and CI checks on every pull request.Enlarge screen
Recurring work on a trigger. Review every pull request, or run maintenance on a schedule.Enlarge screen

Actual Lispa interface · Fictional demo data · Take the full product tour ↓

How your AI team works

From the first idea
to the final review.

Give Lispa the outcome you want. Agents work through the pieces and bring back the changes, checks, and context for your review.

  1. A developer types one brief on a laptop; the Lispa coordinator turns it into three task cards laid out on the desk.
    01 / THE BRIEF

    Start with the outcome.

    Connect your repository and describe the result you want. Lispa turns the brief into tasks and coordinates the agents.

    Enlarge the project screen
  2. The Lispa coordinator holds the plan while Claude, Codex and Grok agents build, test and review at the same time, each on its own cloud machine.
    02 / THE TEAMWORK

    Let the team connect the pieces.

    Agents build, test, and review. Independent tasks move forward together in cloud workspaces that keep running when you close your laptop.

  3. The Lispa coordinator presents a finished pull request with passing checks; the developer presses the accept button.
    03 / YOUR REVIEW

    Open the work.
    Check the evidence.

    Inspect the changes and checks. Ask for a follow-up, request changes, or decide what ships. Your team has the final say.

    Enlarge the review screen

Screens show the actual Lispa interface with demo data. Provider assignments are illustrative.

The teams, together in your project.

One storefront launch. A coordinator follows the plan while build and test agents work in parallel. The review agent brings completed work back to you.

Actual Lispa interface · Fictional project, agents and model assignmentsEnlarge screen

A result you can inspect.
A decision you control.

The same project, one step later: review the payment fix and its checks. Accept the delivery or send it back with changes.

Actual Lispa interface · Fictional result and checksEnlarge screen

Pick something from your backlog.

Reproduce [bug] on [branch]. Add a regression test, fix the cause, and run the relevant checks. Show the diff and test output. Do not merge.

Replace the brackets with your task, then paste the prompt into your first Lispa chat. Judge the result by the code, checks, and recorded usage. Read the first-task guide and review checklist.

Continue with Google

Lite: $20/month, including $20 in usage credits.
Choose your plan after signup, then connect your repo.

Built in Lispa, used in Lispa

The product works
on itself.

This tasks dashboard inside Lispa was built by Lispa agents from a brief like the ones above. The screenshot shows the delivered component with fictional counts: one example of the work an agent brings back for review.

See the feature development workflow →
Built in Lispa · Fictional demo countsEnlarge screen
A little less on your plate

Less chasing.
More making.

AI can write code. Keeping all the work moving is another job. Give that job a home, too.

Separate agents, separate threads
A developer at a desk scratches their head while three separate AI chats each work on the same page, a stack of receipts beside the laptop.

You become the go-between.

  • Repeat the context in another chat.
  • Track down what each agent changed.
  • Piece the results back together.
One goal, a team in Lispa
Claude, Codex and Gemini agents hand completed work to a developer, with an actual Lispa project screenshot inside the illustrated monitor.

You focus on the result.

  • Keep the brief and repository together.
  • Follow agents and their progress in one place.
  • Review the changes and checks, then decide.

Actual Lispa interface · Demo data · View the project screen ↗

Take a closer look at the workspace

Know where the savings come from.

Pay for the model power
each step needs.

The strongest models for hard work. Lower-cost models for routine steps. Lispa can change models as the task changes, keeping the work and context in place.

Choose models yourself, or let Lispa choose for each step.

Savings depend on the work and models used. Fixed subscription prices stay the same. Machines and Lispa services use credits.

Routing off

Opus for every step

$1.45

model calls for nine benchmark tasks

Benchmark score
0.98
Models used
Opus
Run time
Shorter
Lispa routing

Opus for the hard parts

$0.18

model calls for the same nine tasks

Benchmark score
0.98
Models used
Opus + GLM 5.3
Run time
Longer

88% lower model cost in the routing v5 benchmark, with the same checks passed.

Historical internal benchmark · Not a promise about your billMethod and limits ↓
See how model choice changes during a task
One task. More than one model.TRY THE SWITCH ↓

Fix pagination duplicates.

demo/storefront · fix/pagination · same cloud workspace
Your main model
  1. Read the relevant codeOpen model
  2. Find and fix the pagination bugClaude
  3. Run checks and summarizeOpen model
Premium reasoning. Only where it’s needed.

Routine stretches use lower-cost open models. The task stays in place.

Illustrative route. Models and switches depend on your connections, settings, and the work.

01 / KEEP YOUR SUBSCRIPTIONS

Already using Claude Code or Codex?

Give Lispa a separate task while you keep working in your editor. Connect your existing plan with no extra Lispa model-call fee. Your provider’s limits still apply; machines and services use credits.

02 / SPEND LESS ON ROUTINE STEPS

Watching your API bill climb?

Let lower-cost open models handle routine stretches in the same task. Keep Claude or Codex available for harder work. Inspect the model use and estimated cost in the chat.

03 / SHARE ONE BUDGET

Adding agents across a team?

Invite teammates without another seat charge. Share one credit balance for models and machines, and set a credit limit for each member.

Method and limits of the cost exampleWhat was measured, and what it does not show

Historical internal development benchmark, routing v5: nine tasks, one repeat, Opus as the session model and GLM 5.3 at low effort as the balanced tier. Model costs only; Lispa fees and machines are excluded. This is not a current-product benchmark or a promise about quality, speed, or your total bill. Actual savings depend on the task, selected models, cache use, and connected plans. Your results will vary.

The in-product savings comparison prices the same completed tokens at your default model’s API rates. It does not measure a reduction in your subscription bill.

Start with one task. Review the code, checks, and total usage before taking on more work.

See all charges before you start →
A closer look at Lispa

The work has a home.

From the first brief to the next useful lesson. Take a look around the workspace.

Explore the full illustrated walkthrough
Projects

One place for the whole project.

Keep the brief, the work in progress, and the results ready for review together. Give direction without starting the explanation again.

Works with
  • Claude
  • Codex
  • Grok
  • Gemini
  • Open models
Lispa project overview showing the Storefront launch brief and a payment fix ready for review.
Actual Lispa interface · Fictional demo dataView full size
Tasks

Know what needs your attention.

See who is working on each task, which work depends on it, and what is ready for your review. The small pieces stay connected to the goal.

Works with
  • Claude
  • Codex
  • Grok
  • Gemini
  • Open models
Lispa task list grouped into in progress, in review, and done, with a different agent or team assigned to each task.
Actual Lispa interface · Fictional demo dataView full size
Pull requests

Bring the evidence to the review.

Pull requests, agent reviews, and CI checks share a view. Inspect the current change and its latest review before deciding what should ship.

Works with
  • Claude
  • Codex
  • Grok
  • Gemini
  • Open models
Lispa pull request workspace with three changes, the selected payment fix, passed checks, and its latest review.
Actual Lispa interface · Fictional demo dataView full size
Automations

Give recurring work a routine.

Review a new pull request or run a scheduled maintenance pass. See the instructions, the trigger, and the last run in one place.

Works with
  • Claude
  • Codex
  • Grok
  • Gemini
  • Open models
Lispa automations showing a pull request review triggered by GitHub and weekly repository maintenance on a schedule.
Actual Lispa interface · Fictional demo dataView full size
Model routing

Match the model to the moment.

Let Lispa choose model power and reasoning effort for each turn, or keep the model you picked. Change the default or adjust an individual session.

Works with
  • Claude
  • Codex
  • Grok
  • Gemini
  • Open models
Lispa routing settings with automatic model routing enabled and per-session controls explained.
Actual Lispa interface · Fictional demo dataView full size
Shared brain

Leave the next agent a head start.

Keep useful facts, lessons, and reusable commands connected. Explore what the brain holds and how knowledge from one task relates to another.

Works with
  • Claude
  • Codex
  • Grok
  • Gemini
  • Open models
Lispa brain console with checkout facts, lessons, reusable operators, and a graph of their relationships.
Actual Lispa interface · Fictional demo dataView full size
Usage

See where the credits go.

Follow usage over time, session counts, and estimated routing savings. Open the account view to understand the numbers behind the work.

Works with
  • Claude
  • Codex
  • Grok
  • Gemini
  • Open models
Lispa usage chart with fictional demo amounts: what nine October sessions used each day beside what routing saved, and the estimated saving per session.
Actual Lispa interface · Fictional demo dataView full size

Demo amounts only. Routing savings are API-price estimates, not bills or promised savings.

Several agents. One shared goal. Bring back the evidence. Keep your way of working.
Built for developers

The details that matter
when it is your repository.

Isolated cloud machines

Each task runs on its own Linux machine with your repository checked out. Stopped machines keep their state for 30 days.

Branches and worktrees

Agents work on branches in worktrees of your repository, and push or open a pull request only when you say so.

Terminal, files, diff, browser

Follow the work as it happens: open the terminal, read the files and the diff, or watch the browser the agent is testing in.

Your plans and keys

Connect your Claude or ChatGPT plan or a provider API key. Model calls on your own plan carry no Lispa model-call fee.

Limits you control

Set a credit limit for each member, see every charge in Usage, and turn routing on or off for any session.

Skills that travel

Keep reusable skills with the project. Claude Code and Codex sessions read the same skills, so the know-how is not locked in.

Put your AI team to work

What is on your plate?

Start with a real task. See the workflow and a prompt you can make your own.

Explore all use cases →
A familiar piece of work

Your industry.
Your next useful task.

16 industries. 48 starter briefs. Explore the software work Lispa can help your team move forward.

Scroll for more industries · Explore all 16 industries →

Beyond a single chat

A shared brain.
Less repeated work.

Agents create their own reusable language of commands. They save how to solve a problem, so other teams can use it again.

A Codex agent writes a solved recipe into Lispa’s shared brain; Claude and Grok agents pick up the same recipe for their own tasks.
Solve it once. Save it in the shared brain. Every team reuses it.

Put recurring work on a trigger.

Review new PRs, investigate failing CI, or run a weekly maintenance task. Set the instructions and limits. Every run has a chat you can inspect.

CI fails → reproduce → prepare a fix

A language the agents build.

Agents turn repeated steps into commands they can run again. A solved workflow becomes a reusable tool for the next task.

A working solution becomes a reusable command

Knowledge shared across teams.

The project’s brain keeps conventions, debugging lessons, and reusable commands. Connected agent teams can pick up what another team has learned.

Shared project knowledge across sessions
One plan for your organization

Start at $20.
Bring the whole team.

Every plan includes the platform and unlimited members. Credits pay for usage. One credit equals $1.

Lite

Start with a task from your backlog.

$20 / month

20 credits included monthly

Per organization, with unlimited members.

  • Cloud coding agents
  • All models you connect
  • Projects and automations
  • Shared repository knowledge
Start with Lite

Pro

More room for everyday work.

$100 / month

105 credits included monthly

5% more credits than the plan price.

  • Everything in Lite
  • Unlimited members
  • The same per-usage rates
Choose Pro

Max

For more tasks and parallel work.

$200+ / month

220+ credits included monthly

$200–$1,000 in $100 steps. 10% extra credits.

  • Everything in Pro
  • Unlimited members
  • The same per-usage rates
Choose Max

Plans are billed monthly per organization. Cancel any time. USD, applicable taxes extra.

What uses your credits

Models on Lispa’s keys

Provider model cost + 10%. Connected personal plans and keys have no Lispa model-call charge; their own billing and limits apply.

Cloud machines

Standard: 0.53 credits/hour while running. Larger sizes: 1.05 or 2.10 credits/hour. Machines stop when idle.

Routing & knowledge services

Charged by tokens at 20% of the session model’s rates, even with your own model plan. These are usage charges, not a share of savings.

Extra credits

$1 each, $5 minimum. Extra credits carry over; monthly plan credits reset. Auto top-up is off by default.

Connecting a subscription does not make cloud tasks free. Your plan’s credits cover Lispa usage; you can add credits when needed.

Before your first task

A few practical answers.

I already use Claude Code or Codex. Why add Lispa?

Give a cloud agent a separate task while you keep working in your editor. Use your existing plan, switch to cheaper models for routine work, and return to the diff and test output. Lispa adds cloud execution and model choice; it does not reduce the price of your existing subscription.

Do I need my own model subscription or API key?

No. Lispa includes models you can use with your organization’s credits. Connecting your Claude or ChatGPT plan, or a provider API key, gives you more choices. Provider usage limits still apply.

Can I choose the model myself?

Yes. Pick a model, switch mid-task, or turn routing off. Routing chooses among the models available to your workspace; it does not remove your control over the choice.

Where does my code run, and who controls merging?

Cloud tasks run on Linux machines, using branches and worktrees of your repository. You can inspect the terminal, files, diff, and browser. Agents push or merge when you instruct them to; configure automations with the permissions and limits you want.

How much will I save?

Lower-cost models can reduce metered model spend; your fixed subscription price stays the same. Your total also includes cloud machine time and Lispa services. The historical benchmark is one example, not a guaranteed saving. The chat’s savings display compares token costs at API rates; it does not measure subscription-bill savings.

Do I need to know Lisp?

No. Describe the work in plain language. Lispa agents can save reusable operations in Lisp, so future sessions can call them again. You can inspect and edit those operations.

Can I cancel after trying it?

Yes. Cancel from Billing at any time. Your organization keeps access through the paid month. Lite starts at $20 with 20 usage credits.

Prefer working from your terminal?The Lispa agent also runs on macOS and Linux.
curl -fsSL https://lispa.ai/install.sh | sh
For investors

A business built on AI savings.

Lispa is raising a pre-seed round. Founder Tedj Meabiou, based in Dubai, will send the deck and walk you through the product.

Request the deck & a demo

tedj@lispa.com · +971 55 189 7923

One repository. One useful task.

Give your AI team
its first goal.

Bring a bug, a test gap, or a small feature.
Review the result and its cost before committing to more.

Get startedContinue with Google

Create your account. Choose a plan. Connect your repo.
$20/month includes $20 in usage credits · Cancel any time

More signup options
Product screenActual Lispa interface · Fictional data · Scroll to inspect