Connect what you choose
You decide which tools SylvieSI can read, and what it may do in each.
Tools SylvieSI connects to
SylvieSI reads user messages, code and error data from the tools you already use, and sends updates back through them. Each connection has its own permissions.
Permissions
You choose which tools to connect and what SylvieSI may do in each. You can disconnect any tool at any time.
ToolWhat SylvieSI does with itAccess it needs
In-app messages: reports from inside your product
Receives messages that senders write from inside your product.
Your server signs each message with the sender’s identity.
Gmail: a support inbox SylvieSI can read and answer
Reads a support inbox you choose and replies on the same thread.
Google authorization for that inbox. You can limit it by label and choose how far back to read.
Slack: bug reports and replies in your channels
Reads the support channels you choose and replies in the thread.
The SylvieSI Slack app, installed on the channels you pick.
GitHub: code reading, branches and draft pull requests
Reads your code to find the cause of a bug. Creates branches and draft pull requests on repositories you approve.
Read access to the repository. Write access to branches and draft pull requests only after approval. SylvieSI never merges.
Linear: engineering issues with a full handoff
Receives engineering issues with the full handoff and syncs their status back to the Case.
Linear authorization to create and read issues.
Sentry: errors and traces for the cause
Reads errors and traces to show how your product behaves when a bug occurs.
Sentry authorization for the projects you choose.
Model provider: the model behind each agent
Runs the agents. The model for each role is a setting.
An API key from the provider you pick. Secrets are never sent to a model.
Every connection shows its state
A connection is Not connected, Connecting, Connected, Degraded, Expired or Revoked. Degraded shows the reason, such as missing data. When a source is out of date, SylvieSI says so and lowers its certainty about the cause.
See what is missing
A disconnected or out-of-date tool is shown, not hidden.
Confirm every action
An action counts as done only when the tool confirms it.
Manage each connection
See the permissions, who authorized it and the time of the last health check. Re-test or disconnect from the same page.
Reading and writing are approved separately
Each connection is approved for a specific tool and a specific action.
Read access
Which account, workspace, repository or environment SylvieSI may read, and how fresh that data is.
Write access
The action, its target and the approval it needs. Being able to read an issue does not allow SylvieSI to change it.
If a permission is revoked or a tool fails, SylvieSI shows it and stops the work that depends on it.
Read about authorization