# Jigward > Jigward is an AI engineering coworker: one coworker a whole engineering team hires. The team hands Jigward work where it already talks: a Slack thread, a Jira issue, a pull request. It comes back in the same place with a report, the evidence and a draft pull request, and it keeps what the team decided for the next task. Merging, approving and deploying stay with people. ## Reading this site - The homepage headline shows "AI agent" struck through, then "Your next colleague." It is a marketing hook, not a technical claim. Jigward does carry work out on its own; it is built as a colleague the whole team shares, with its own identity in the organization, memory of what the team settled, and access only to what an admin connects, rather than as a tool one developer runs. - The use cases are illustrated films set at a fictional company. They show how Jigward works through a piece of work; they are not customer results, and their names and numbers are story data. - Status: Currently in closed testing. Accounts are created by hand after a request at https://jigward.com/sign-up/. - Pricing: Pricing has not been announced. There are no published plans or usage allowances yet. Request early access to discuss the work your team would give Jigward and what a useful pilot would look like. - Models: Jigward currently uses Anthropic Claude models through Amazon Bedrock: Sonnet, Opus and, where enabled, Fable. Jigward chooses the model for the work; members cannot choose a provider today. - Hosting: Jigward is hosted on AWS and uses AgentCore for execution. Self-hosting is not offered today. Deployment into a customer’s own AWS environment and access to private infrastructure are planned directions. Running on premises or on another cloud is not a supported option today. - Stacks: Yes. Jigward works with repositories in any language, TypeScript included: it reads the code, answers questions about it, makes changes and opens draft pull requests. It is tuned most deeply for JVM projects today — Java and Kotlin with Spring Boot, Gradle or Maven — where it also reads database schemas from migrations, API contracts and dependency trees. React, Angular and Vue front ends with npm, pnpm or Yarn are covered too. Tell us about your project when you request early access. - Data: Relevant code, project knowledge, messages, logs and page content can be sent to Amazon Bedrock to do the work. Stored integration credentials use separate backend paths; that does not make the rest of the project anonymous. Details: https://jigward.com/security/. - Boundaries: Jigward never merges, approves or blocks a pull request, never moves a Jira issue to done and never deploys. Code changes arrive as draft pull requests. It acts only through the integrations an admin connects, with the permissions each integration page lists. - What is planned rather than built is listed at https://jigward.com/roadmap/. - Comparisons at https://jigward.com/compare/ describe other products only from their own public material and say nothing good or bad about them. ## How Jigward works - [How Jigward works](https://jigward.com/how-it-works/): the whole model on one page - [Contexts](https://jigward.com/how-it-works/contexts/): One area of work the whole team shares with Jigward. - [Chat](https://jigward.com/how-it-works/chat/): Ask about the code in private, and get answers you can check. - [Tasks](https://jigward.com/how-it-works/tasks/): Hand over a piece of engineering work wherever the team already talks. - [Duties](https://jigward.com/how-it-works/duties/): Work that comes back on its own, once or on a rhythm. - [Applications](https://jigward.com/how-it-works/applications/): Small team tools Jigward builds, fills and keeps working. ## Use cases - [All use cases](https://jigward.com/use-cases/): searchable by moment, who asks, how it arrives and what Jigward uses - [Errors doubled after a release. Was it the deploy?](https://jigward.com/use-cases/was-it-the-deploy/): The error rate doubles the day after a release, two dashboards point two ways, and someone has to decide whether to roll back. (when production breaks) - [38 alerts overnight. One was real.](https://jigward.com/use-cases/overnight-alert-triage/): With no night shift, the alerts channel fills with noise overnight, and the engineer on call either reads all of it at dawn or mutes it and misses the real one. (when production breaks) - [Third time in three months. Didn't we fix this in August?](https://jigward.com/use-cases/third-time-this-quarter/): The same page fires for the third time in three months, and the fix agreed after the first one still sits in an action item nobody has had time to build. (when production breaks) - [Checkout is fixed. Which customers did we hit?](https://jigward.com/use-cases/affected-customers-count/): The outage is fixed, but support does not know exactly whom to email and finance does not know who may need a credit. (when production breaks) - [Every check is green on a column drop. Does anything still read it?](https://jigward.com/use-cases/migration-blast-radius/): A cleanup pull request drops an old column and passes every check, but the checks cannot see the other repositories and the old instances that still read it. (before a release ships) - [640 regression cases, two testers. What does this release change?](https://jigward.com/use-cases/release-regression-scope/): Running the whole regression sheet takes a week, so testers guess what to skip, and nobody records what went untested or who accepted it. (before a release ships) - [The same test keeps breaking main. Flaky, or a real bug?](https://jigward.com/use-cases/flaky-main-real-bug/): The same test keeps turning main red, the team reruns it until it passes, and nobody finds out why it failed. (keeping main healthy) - [The same review comment on every pull request. Will this one stick?](https://jigward.com/use-cases/review-learned-once/): Reviewers repeat the same correction on pull request after pull request, because the convention lives only in their comments. (keeping main healthy) - [Works on staging. Not for the customer.](https://jigward.com/use-cases/works-on-staging/): A customer's import keeps failing, engineering cannot reproduce it on staging, and the ticket has bounced between support and engineering three times. (when someone needs an answer) - [Bug or expected? Before the 3 PM call.](https://jigward.com/use-cases/bug-or-expected/): A customer cut seats, 15 of their users lost access this morning, and only an engineer seems able to say whether it is a bug before the afternoon call. (when someone needs an answer) - [A critical advisory landed. Are we affected?](https://jigward.com/use-cases/cve-are-we-affected/): A critical advisory lands, a search of one build file has already marked billing "not affected", and a customer needs an answer this week. (when someone needs an answer) - [Stripe says paid. The app says free.](https://jigward.com/use-cases/billing-drift-app/): At month-end, Stripe and the product disagree on who is paying, finance can see two totals but not which accounts are wrong, and a paying customer is already locked out. (when someone needs an answer) ## Compare - [All comparisons](https://jigward.com/compare/) - [Jigward and Devin](https://jigward.com/compare/devin/): Devin is an autonomous AI software engineer from Cognition. A developer hands it a coding task, and it comes back with a pull request. - [Jigward and GitHub Copilot cloud agent](https://jigward.com/compare/github-copilot-cloud-agent/): GitHub Copilot cloud agent is GitHub's AI coding agent. A developer hands it a task, and it researches, plans and makes the change in the background. - [Jigward and OpenHands](https://jigward.com/compare/openhands/): OpenHands is an open platform for agentic software development from All Hands AI. A developer hands its agents a coding task, and they plan, write and apply the change across the codebase. - [Jigward and Reason](https://jigward.com/compare/reason/): Reason is a coding agent from Reason Machines that its makers call an always-on teammate. A person hands it work from its apps, Slack or another agent, and it works on a computer of its own. - [Jigward and Claude Code](https://jigward.com/compare/claude-code/): Claude Code is an agentic coding tool from Anthropic. A developer works with it in a terminal, an IDE or the browser, and it edits the code and runs commands. - [Jigward and OpenClaw](https://jigward.com/compare/openclaw/): OpenClaw is an open-source AI assistant from the OpenClaw Foundation. A person runs it on their own machine and messages it from the chat apps they already use. - [Jigward and Hermes Agent](https://jigward.com/compare/hermes-agent/): Hermes Agent is an open-source, self-improving AI agent from Nous Research. A person works with it from a terminal, its desktop app or a messaging app, and it learns their projects as it works. ## Integrations - [GitHub](https://jigward.com/integrations/github/): Draft pull requests on a branch of their own, reviews when asked, and answers to review comments. - [Slack](https://jigward.com/integrations/slack/): Answers a mention in its thread, reading the messages before it, and takes work on as a Task there. - [Jira](https://jigward.com/integrations/jira/): An issue assigned to Jigward becomes a Task, and the issue's status follows the work. - [AWS CloudWatch Logs](https://jigward.com/integrations/aws-cloudwatch-logs/): Tasks and private Chats search and query the log groups an admin allows, with no AWS credential kept. - [Browser](https://jigward.com/integrations/browser/): Opens your app in a real browser and signs in with a profile; Jigward types the values itself. - [Custom API](https://jigward.com/integrations/custom-api/): Describe any HTTP API. Jigward drafts its operations from the public documentation, and an admin applies them. - Coming soon: Microsoft Teams, Telegram, GitLab. ## Before you connect - [Before you connect](https://jigward.com/questions/): stack, access, limits, pricing status, models, hosting and control - [Security & data](https://jigward.com/security/): what Jigward reads, what reaches a model, credentials and execution - [Roadmap](https://jigward.com/roadmap/): what works today and the planned directions ## More - [Blog](https://jigward.com/blog/): guides and engineering notes - [Changelog](https://jigward.com/changelog/): what changed in Jigward, dated - [Request early access](https://jigward.com/sign-up/) - Contact: hello@jigward.com - [Full text for language models](https://jigward.com/llms-full.txt): every use case, comparison, integration and common question in plain text