SylvieSI is made by THE SHRIKS. It has three parts. Mobius collects and sorts what users send. Ouroboros investigates bugs and prepares changes. Sylvie is the workspace where your team directs both. Ten AI agents do the reading and drafting, and a person approves every change.
Explore the three parts ↓Sorts and groups
SYLVIEDirects the work
OUROBOROSFixes or builds
Each part has its own job. All three work on the same bugs and requests.
Sylvie is where your team works with both of the other parts. You write a goal in plain words, and Sylvie plans the steps, gathers findings and keeps the goal open after the chat ends. You can ask a question, start an investigation, prepare an action or schedule a task.
Sylvie proposes and explains. It does not carry out an action by itself.
Explore Sylvie ↗Mobius is the part of SylvieSI that listens. It reads messages from your app, Slack and Gmail, works out who wrote each one, labels it as a bug, a feature request, a question or feedback, and groups messages about the same thing. Repeated requests become an Idea that lists the arguments for and against.
Mobius never changes your product. It organizes what users say and keeps every original message.
Explore Mobius ↗
Ouroboros is the part that does the engineering work. It takes a confirmed bug or an approved feature request and follows it to a checked result. It reproduces the bug, finds the cause, prepares the change, asks a person to approve it, tests it on staging and saves the outcome.
Ouroboros never merges a pull request and never deploys on its own. Nothing changes until a person approves the exact change.
Explore Ouroboros ↗Each part has a place of its own, and a name that goes with it.
Sylvie is the research vessel. Your team works from it: you set a goal, and the ship plans the route, gathers the findings and reports back.
Mobius is the lab. It watches everything users send, sorts it, groups duplicates and keeps every original message.
Ouroboros, or O.B., is the engineering chamber. It takes a confirmed bug or an approved request and carries it to a checked result.
The names fit the work. A bug that returns re-opens its Case, and that Case feeds the next investigation, so Ouroboros keeps learning from its own past. SylvieSI is the whole: the ship, the observatory and the chamber, made by THE SHRIKS.
Agents read, investigate and propose. A controlled service carries out any action after the approval policy is met.
Plans an investigation or goal, chooses which agents to run and combines their findings into one cited answer.
Answers questions about the item you have open, using only that item and the sources linked to it.
Labels each message, suggests duplicates and drafts the question to ask when information is missing.
Reads connected sources such as code and error data, and saves each finding with its source.
Gathers feature requests into an Idea and lists the arguments for and against building it.
Forms hypotheses about a bug and writes a root cause assessment with supporting, contradicting and missing evidence.
Matches an approved decision to a component in the approved catalog and prepares it for staging.
Writes the handoff, reproduction steps and test plan, and prepares a branch and a draft pull request.
Drafts replies to senders and never claims a result that has not been checked.
Writes the overview summaries of what changed, with sources.
In Settings, you turn each agent on or off, choose the model behind each role and set spending limits.
Each step hands the next one a specific item.
Bug reports and feature requests, grouped, each with the sender’s own words.
A confirmed bug with a root cause assessment, or a feature request with the arguments for and against.
A Linear issue or a draft pull request, or an approved component for staging.
A person’s decision on the exact change, with its risk and how to undo it.
The checks that passed, where they ran, and the reply sent to the sender.
SylvieSI has three modes on the same items, and a set of tools around them.
Customer Intelligence has an overview, an inbox, a Customers list and Insights with Ideas. Product Operations has an overview, Cases, Work, Changes, Approvals and Outcomes. Sylvie has Chats, Objectives and Connectors, with a side panel for running work, approvals and schedules.
Five steps cover your account, workspace, message sources, product details and permissions.
Seven roles: Owner, Admin, Product, Engineering, Support, Reviewer and Viewer.
One search covers messages, customers, cases, ideas, work, changes, approvals, objectives and chats, and follows your role.
Twelve settings pages cover Workspace, Team, Product and Environments, Product Graph, Runtime, Integrations, Agents, Automation and Policies, Permissions, Data and Privacy, Notifications and Audit.
Explore Sylvie ↗See every featureThe Product Graph holds what SylvieSI knows about your product: its pages, components, APIs, roles and processes, read from your code and API description.
Links a bug or request to the screens and APIs it affects.
Each fact shows where it came from and how old it is.
A person can correct a fact. The correction is saved as a new version and the old one stays.
An investigation can start from earlier Cases in your workspace that had the same cause.
Agents read, investigate and propose. A person approves the exact change before anything runs. Nothing merges or deploys on its own.
Tell us about your product and where user messages arrive today.
Request access ↗