Skip to main content
Get Reliant running, chat with your codebase, and run your first workflow.

​
1. Connect an AI provider

Open Settings → AI.
  1. Select your provider: Anthropic, OpenAI, OpenRouter, Codex, and others.
  2. Authenticate:
    • Most providers: enter an API key, then click Test Connection and Add Provider.
    • Codex provider: click Login with Codex and complete OAuth in your browser.

​
2. Open a project

Click Open Existing Project and select your code folder. Reliant indexes your files automatically. Or click Create New Project to start fresh.

​
3. Start chatting

Type a message and press Enter:
  • “Explain the structure of this project”
  • “What does the main file do?”
  • “Add input validation to the signup form”
This is the default agent workflow running. Your message goes to the LLM, the LLM calls tools (read files, search code, make edits, run commands), results go back to the LLM, and it loops until the task is done. So far this is like any other ADE.

​
4. Run a built-in workflow

This is where Reliant gets different. Instead of manually driving each step, pick a workflow and let agents execute a defined process. Click the workflow selector (defaults to “Agent”) and try one of these:
WorkflowWhen to use itWhat happens
gsdYou have a clear feature to buildDiscuss → research → plan → parallel implementation → verify
superpowersYou want test-driven developmentBrainstorm → plan → write failing tests → implement → review
simplify-firstYou’re working in messy codeAnalyze simplifications → refactor → verify → then build
spec-drivenYou want to nail requirements firstWrite spec → plan approach → implement
get-it-rightComplex change in a large codebaseAttempt → evaluate → improve/restart → diagnose → final implementation
Select a workflow, type your request, and watch multiple agents collaborate on the task.

​
5. Build your own workflow

Create a YAML file in your project to define a custom process:
.reliant/workflows/review-and-refactor.yaml
name: review-and-refactor
version: v0.0.1
description: Reviews code, then refactors based on findings
status: published
tag: agent

entry: [review]

inputs:
  model:
    type: model
    default: ""

nodes:
  - id: review
    workflow: builtin://agent
    thread:
      mode: inherit
    args:
      model: "{{inputs.model}}"
      mode: auto
      system_prompt: |
        Critically review the code the user points you to. Focus on:
        - Unnecessary complexity
        - Duplicated logic
        - Unclear naming or structure
        Be specific. Reference exact code.

  - id: refactor
    workflow: builtin://agent
    thread:
      mode: inherit
      inject:
        role: user
        content: |
          Based on your review above, refactor the code to address the
          issues you found. Make minimal, focused changes.
    args:
      model: "{{inputs.model}}"
      mode: auto

  - id: verify
    run: npm test

edges:
  - from: review
    default: refactor
  - from: refactor
    default: verify
This workflow: criticizes the code → refactors based on its own critique → runs tests to verify nothing broke. Three steps, zero manual intervention. See Creating Custom Workflows for the full guide.

​
The interface

AreaPurpose
Navigation BarChats, Worktrees, Workflows, Settings
SidebarContext for the selected navigation item
Chat AreaMessage input at bottom, history above

​
Execution modes

Control how much autonomy the agent has:
ModeBehavior
AutoExecutes tools without asking. Press Escape to cancel.
ManualAsks for approval before each tool execution.
PlanRead-only tools only — explores but doesn’t change anything.
A common pattern: start in Plan mode to analyze and create a plan, then switch to Auto to execute it.

​
Optional: memory files

Memory files provide persistent context to the AI across conversations. Global: ~/.reliant/reliant.md 
# My Guidelines
- Always write tests
- Use descriptive variable names
Project-specific: reliant.md in your project root 
# Project Context
- TypeScript/React project
- Uses Prisma for database
Learn more: Memories & Context

​
Next steps