An AI software engineer for the work around the code.
It reads the repo, the issues and the pull requests, triages the bug reports, runs scripts for the data and ops jobs, and tells you when the build breaks. It doesn't change your codebase: code changes go to an engineer, written up and ready.
300 free credits to start. Live in two minutes.
What is an AI software engineer?
An AI software engineer, in Cicada's sense, is an agent that takes the work around the code off your engineers rather than writing the code itself. It reads your repository, issues and pull requests through GitHub, turns incoming bug reports into issues in Linear or Jira with what an engineer needs, writes and runs scripts in its own sandbox for data pulls and one-off jobs, drafts release notes from what merged, and checks your CI on a schedule. It is not a coding agent for your codebase: changes to the code go to a person.
What your AI software engineer does.
Reads the repo and the issues
Through GitHub: code, issues, pull requests and their discussions. Ask where something is handled, what changed last week or who reviewed a fix, and it reads before it answers.
Triages bug reports
Reads what comes in from Zendesk, Intercom or a Slack channel, groups the duplicates, notes when each started and writes the issue in Linear, Jira or GitHub with steps to reproduce.
Runs scripts in its own sandbox
Data pulls, CSV clean-ups and one-off checks against an export: it writes the script, runs it in its sandbox, not on your servers, and hands back the result and the code.
Drafts release notes and docs
From what merged since the last tag: customer-facing release notes, an internal changelog, or an update to a doc in Notion or Google Docs, for your review.
Watches CI
Checks your GitHub Actions runs on a schedule and messages you when main goes red, with the job that failed and the pull request that merged before it.
Hands code changes to a person
When the fix is a change to the code, it writes it up: what's wrong, where it likely lives, what it would try. An engineer makes the change.
Who hires an AI software engineer.
Triage before standup
Overnight bug reports grouped, deduplicated and written up in Linear before the team logs in.
Ops work off your plate
The data pull, the CSV fix and the release notes done in the sandbox or drafted, so you stay on the product.
Bug reports engineers can use
Every report turned into an issue with steps, the customer's own words and links to the tickets it came from.
Know what shipped
A weekly summary of what merged, what's in review and what's blocked, from GitHub and Linear.
Someone watching the build
A message when main breaks, with the job that failed, wherever you are.
Scripts without a ticket
One-off analysis on an export, run in its sandbox, with the code handed back so it can be rerun.
Set up in three steps.
Say what the job is
Describe the role in a sentence when you sign up, or tell the agent in the chat. It writes its own brief, which you can read and change any time.
Connect the apps it needs
One click for most apps. Each one can be Full, Ask first or Read-only, for every agent in the workspace.
Link your phone, or don't
Scan a code and it's in your iMessage or WhatsApp. Or keep it in the app, Slack and email. Hand it the first job in the same message.
Nothing goes out without your yes.
Nothing goes out without your yes
Anything that can't be undone, and any message to another person, waits for your OK. Approve with a 👍 in the chat, or set a role to full autonomy once you trust it.
Read-only is a real lock
Make any app read-only and the agent can look but never change a thing. It applies to every agent in the workspace, from their next run.
Every step on the record
Each run shows what it read, what it did and what it sent, step by step, in the Activity tab. Corrections you make stick.
What it won't do.
Won't pretend
If it couldn't confirm something, it leaves it out and says so. A number without a source doesn't go in a message.
Won't go around you
It never messages anyone on your behalf unless you asked, never spends money, and never makes a judgement call about money on its own.
Won't touch your codebase
It reads code and writes scripts in its own sandbox. Changes to your repo, merges and deploys go to an engineer.
Three ways to handle the work around the code
| Task | Doing it yourself | A coding assistant in your editor | An AI software engineer |
|---|---|---|---|
| Writing the code | You | Suggests and edits code in your editor | Not covered: it writes up the change for an engineer |
| Triaging bug reports | Whoever reads support first | Not covered | Grouped, deduplicated and written up in Linear or Jira |
| One-off scripts and data pulls | An afternoon | Helps you write it | Written and run in its sandbox, result and code handed back |
| Release notes | Scroll the merged pull requests | If you ask in the editor | Drafted from what merged since the last tag |
| Watching CI | Refresh the Actions tab | Not covered | Checked on a schedule, a message when main goes red |
| Where it happens | Your laptop | Your editor | Slack, iMessage, WhatsApp or email |
Things to ask your AI software engineer.
- 01
"what changed in the payments code since the last release? plain english"
You get: A short summary of each merged pull request that touched payments, who wrote it and what it changed.
- 02
"Triage this week's bug reports from support and put them in Linear."
You get: Duplicates grouped, one issue per real bug with steps to reproduce and links to the tickets, waiting for your OK.
- 03
"Here's the orders export. Find customers who ordered twice in a week, and give me the script too."
You get: A sheet of the matches and the script it ran in its sandbox, so you can run it again next month.
- 04
"Draft the release notes for v2.14 from what merged."
You get: Customer-facing notes and an internal changelog in a Google Doc, grouped into features, fixes and changes.
- 05
"Check CI every morning at 8 and tell me if main is red."
You get: A message on the minute when a run has failed, with the job and the last pull request merged before it.
Frequently asked questions.
Does it write code in our repo?
No. It reads your code, issues and pull requests through GitHub and writes scripts that run in its own sandbox. Changes to your codebase go to an engineer, with the issue written up. Set GitHub to Read-only and it can't change anything there at all.
Where do I talk to it?
Slack, iMessage, WhatsApp, email or the web app. In a channel like #eng it stays quiet until someone names it, then answers in the thread.
What can its sandbox do?
Run Python and Node scripts on files you send it or exports in Drive: pull, clean, join and chart. It hands back the result and the code. It doesn't run on your servers.
How does it triage bug reports?
It reads reports where they arrive (Zendesk, Intercom, a Slack channel), groups the ones that describe the same problem and writes one issue in Linear, Jira or GitHub with the steps, the customers affected and links back.
Can it watch our CI?
Yes. It checks GitHub Actions on the schedule you set and messages you when a run fails, with the job and what merged before it.
What won't it do?
It isn't built to change your codebase: no pushes, no merges, no deploys. Make GitHub read-only and that's a lock, not a promise. It also won't message anyone unless you asked, and won't state a cause it couldn't confirm.
Can the whole engineering team use it?
Yes. Everyone in your workspace gets their own agent. GitHub, Linear and Slack are connected once for the team; personal apps stay under each person's own login.
What does it cost?
Every workspace starts with 300 free credits. You add a card at sign-up and nothing is charged until you run low; then it's $25 for 200 credits, with no subscription. A quick reply is about 5 credits and a typical run under 20; plans start at $99 a month.
Other roles you can hire for
- AI Executive Assistant
- AI Chief of Staff
- AI Virtual Assistant
- AI Personal Assistant
- AI Office Manager
- AI Receptionist
- AI Scheduling Assistant
- AI SDR
- AI Sales Assistant
- AI Account Manager
- AI Customer Success Manager
- AI Head of Growth
- AI RevOps Analyst
- AI Lead Researcher
- AI Marketing Manager
- AI Content Writer
- AI SEO Specialist
- AI Social Media Manager
- AI Community Manager
- AI Content Creator
- AI Email Marketing Manager
- AI Operations Manager
- AI Project Manager
- AI Procurement Assistant
- AI Inventory Manager
- AI Supply Chain Analyst
- AI Ecommerce Manager
- AI Property Management Assistant
- AI Bookkeeper
- AI Finance Analyst
- AI Recruiter
- AI HR Assistant
- AI Data Analyst
- AI Research Assistant
- AI Customer Support Agent
- AI Legal Assistant
- AI Designer
- AI IT Support Assistant
- AI Compliance Assistant
- For Shopify stores
- For home services
- For real estate teams
- For recruiting agencies
- For manufacturers
- For agencies
- For law firms
- For restaurants and hospitality
- For property managers
- For consultants
One teammate. The output of a team.
Works nights, remembers every decision, answers in the thread. Start free with 300 credits.