Repository memory
The CLI does not start from an empty prompt. It can use the repo graph, prior review outcomes, dismissed findings, naming decisions, and code ownership signals that already live in your Sigilix workspace.
CLI
Run Sigilix from your repo.
The CLI puts the Sigilix moat where you work: local repair and review follow-up backed by repository memory, issue context, and the model route you choose.
It is not just another terminal assistant loop. Sigilix connects your codebase, review history, Linear issues, Slack decisions, and CLI sessions so the model can reason with the system around the task.
In the terminal
Review and repair, without leaving your repo.
Local repair and review follow-up from your terminal — backed by repository memory, issue context, and the model route you choose.
Moat
Context that compounds inside your workspace.
Workflow context
Linear issues, GitHub PR history, Slack decisions, and CLI sessions become useful context for the next task. The point is to help you avoid restating the same project facts every time you ask for a repair.
Evidence before answers
Sigilix is built around grounded repair. The model should point to the file, failure path, prior decision, or ticket trail that shaped the answer instead of making the fix sound plausible without proof.
Bounded learning
Memory improves the experience inside your workspace. It is not a product-wide training claim, and private code is not used to train shared models unless you explicitly enable the relevant setting.
Setup
From account to first repair.
Install is tied to your workspace so the CLI can use the same repositories, memory, model routing, and plan permissions you set up in Sigilix. The app gives you a workspace-scoped command instead of a generic public install path.
01
Create an account
Start with a Sigilix account so the CLI can attach local sessions to the right workspace, repositories, plans, and model routes.
02
Connect the workspace
Connect GitHub first, then Linear or Jira if you want issue triage and repair loops. Slack can be added when you want team decisions to become searchable context.
03
Install the CLI
Copy the install command from the Sigilix app after your workspace is connected. The command is scoped to your org, repositories, and plan permissions.
04
Choose the route
Use Sigilix models by default. Bring your provider key for local CLI sessions if you'd like.
Install flow
Create your Sigilix account.
Connect GitHub and the repositories the CLI can use.
Open the Sigilix app and copy the workspace CLI install command.
Sign in from the repo and choose the default model route.
Models
Available routes.
The CLI starts on Sigilix models. You can switch route depth based on the job: fast local help, everyday engineering work, deeper repair.
01 / Light
Boreas 3.1
Fast, repo-aware help for routine CLI questions, small repairs, and everyday review follow-up. This is the default route when speed matters.
02 / Base
Pyroeis 3
The everyday route for product work: multi-file edits, small refactors, issue follow-up, and questions that need project context without a frontier-level spend.
03 / Premium
Astraeus 3
The high-effort route for hard bugs, security-sensitive repairs, architecture changes, and investigations where missed context is expensive.
Plans and CLI BYOK
Use Sigilix models by default. Bring your provider key for local CLI sessions if you'd like.
Free
Includes PR review, CLI access, and research-based questions with a low cap window. Good for connecting a repo and seeing the grounded workflow.
Pro
Adds the paid workflow surface: triage, Slack assistant, larger usage windows, and optional CLI BYOK across Pro, Max, and Ultra. Use Sigilix models by default. Bring your provider key for local CLI sessions if you'd like.
CLI BYOK
Bring your own provider key for local CLI sessions on paid tiers. Provider usage stays on your account; Sigilix still supplies the context layer, routing, and workspace memory.
Research questions
Deep-Research Chat runs on Sigilix models only. BYOK is for local CLI sessions; hosted browser workflows stay governed by the Sigilix model line.
Build with memory. Keep the boundary visible.
Sigilix improves by carrying the right context forward inside your workspace: the bugs you fixed, the reviews you trusted, the tickets you assigned, and the decisions your team already made.
Workspace memory is scoped to your organization and user settings. Private code is not used to train shared models unless you explicitly enable that setting. Review the privacy policy and terms for the current account and data controls.