Skip to main content

Quickstart

This guide gets EstaCoda from a clean machine to a verified local session. Install it, run onboarding, configure one working provider/model setup, verify readiness, and start the CLI before adding channels, voice, browser, image generation, or other runtime surfaces.

Who this is for

Use this page if you are:

  • Installing EstaCoda for the first time
  • Setting up one provider/model pair
  • Verifying the CLI before enabling Telegram, WhatsApp, voice, browser, or image generation
  • Recovering from an install that completed but does not produce a working session
  • Preparing a source checkout for development

The fastest path

Pick the row that matches your goal:

GoalStart hereThen
I want EstaCoda working locallyInstall, then run estacoda setupAccept the launch prompt or run estacoda
I already installed itestacoda verifyCheck estacoda model status
I ran estacoda and setup appearsContinue the setup promptUse estacoda setup if you exit early
I want Telegram or WhatsAppGet one local CLI session working firstConfigure channels after verification
I want voice, Search, browser, or image generationFinish the base setup firstAdd optional capabilities one at a time
I want to modify the sourceClone the repoRun ./scripts/setup-estacoda.sh
Something feels brokenestacoda doctorFix the first blocker before adding features

If a normal local CLI session does not work, do not enable channels, voice, browser, image generation, or evolution surfaces yet. First get one clean local turn working.


1. Install EstaCoda

The default install path uses the public installer:

curl -fsSL https://www.estacoda.com/install.sh | bash

This creates a managed-source install under ~/.estacoda/estacoda, builds the project, writes a wrapper to ~/.local/bin/estacoda, and runs estacoda init.

If ~/.local/bin is not on your PATH, add it:

export PATH="$HOME/.local/bin:$PATH"

2. Run first-run onboarding

Start the reviewed setup flow:

estacoda setup

If you run bare estacoda before setup is complete, EstaCoda detects the incomplete state and guides you back to setup.

The Onboarding Wizard walks through the setup in this order:

StepWhat it doesResult
Language and interface styleChooses setup language and terminal styleSetup uses the selected UI mode
WorkspaceSelects the workspace pathEstaCoda knows where it is allowed to operate
Workspace trustSets trust for the selected directoryTool behavior follows the trust decision
Provider endpoint and modelChooses the primary provider/model setup, including local/custom endpoint URL when neededA usable model configuration is ready
Credential handlingStores, reuses, or skips credentials depending on provider authSecrets remain out of review output
Security modeChooses approval behaviorRisky actions require the configured approval level
Agent EvolutionChooses learning behaviorEvolution behavior follows review settings
Optional capabilitiesChannels, voice, Search, and browser setupExtra surfaces are configured only when selected
Review and applyShows a final summary before writingChanges are applied after confirmation
Verify and launchChecks readiness and offers launchYou reach a local session

Setup is reviewed before apply. The wizard shows a summary before committing configuration changes, and raw secrets are not displayed in review output.

The wizard includes Back on choice prompts where it helps correct a prior setup choice. Text and secret entry prompts keep their normal behavior.

First-run onboarding can configure Channels, Voice, Search, and Browser. Image generation, fallback models, and deeper provider configuration are available after the base setup through the Setup Editor and dedicated CLI commands.

3. Configure one working model

Choose one working provider/model pair first. For local or custom OpenAI-compatible servers such as Ollama, LM Studio, llama.cpp, or vLLM, select the Local / Custom Endpoint path and enter the endpoint URL when prompted. API keys are optional for that path. Do not configure fallback models, channels, or optional capabilities until the base setup verifies.

After setup, inspect the active model configuration:

estacoda model status

Profile state is written under:

~/.estacoda/profiles/default/

4. Start your first session

If the setup wizard offers to launch EstaCoda, accept it. Otherwise run:

estacoda

Use a prompt that is easy to verify:

Check this directory and tell me what kind of project this is.
Then list the first three files you would inspect next.

A working first session means:

CheckWhat success looks like
StartupEstaCoda starts without provider or model errors
Model configurationestacoda model status shows the selected setup
First answerThe agent responds normally
Second turnA follow-up works in the same session
Tools/tools shows available tools or explains why tools are unavailable
SafetyApprovals or denials explain what happened

5. Verify setup

Run these after the first session:

estacoda verify
estacoda model status
estacoda doctor

Use estacoda doctor --live only when the local setup looks correct but provider connectivity still fails.

6. Try the core controls

Inside a session:

CommandUse
/helpShow available session commands
/toolsList available tools
/modelInspect or change the session model configuration
/statusShow current session/runtime status
/interruptCancel an active turn
/exitLeave the session

Type / to open the slash menu. Use /help inside a session to see the active controls for your installed version.

7. Add the next layer

Only add optional surfaces after the local CLI works.

Next layerUse whenStart with
TelegramYou want remote access from TelegramChannels
WhatsAppYou want QR-linked WhatsApp accessChannels
VoiceYou want STT/TTSVoice
SearchYou want live web search from Brave or DDGSProvider Reference
BrowserYou want browser-backed workBrowser
Image generationYou want provider-backed image toolsImage Generation
SkillsYou want reusable task proceduresSkills
MCPYou want external tool serversTools

8. Common failure modes

SymptomLikely causeFirst fix
estacoda opens setup instead of a sessionSetup is incompleteRun estacoda setup
Setup completes but launch is blockedWorkspace trust was deferredRe-run setup or trust the workspace
A model is configured but cannot answerMissing or invalid credentialRun estacoda model status, then estacoda setup
Commands are missing after install~/.local/bin is not on PATHAdd it to your shell profile and reload
Tools are unavailableTool policy or profile config does not expose themRun /tools and estacoda verify
Channel messages do not arriveChannel auth, allowlist, or platform setup is incompleteCheck channel docs and diagnostics
Behavior changed after editsProfile config or trust state changedRun estacoda doctor before changing more settings

9. Recovery order

Use this order before manual debugging:

estacoda verify
estacoda model status
estacoda doctor
estacoda setup

Use estacoda doctor --live after that only if provider connectivity is still the suspected failure.

10. Advanced install paths

Install to a custom directory and skip the initial state bootstrap:

curl -fsSL https://www.estacoda.com/install.sh | bash -s -- --dir <path> --skip-init

If you plan to modify the source, clone the repo and run the setup script:

git clone https://github.com/sifr01-labs/EstaCoda.git
cd EstaCoda
./scripts/setup-estacoda.sh

This creates a manual-source install. The checkout is preserved during uninstall, and update operates in check-and-advise mode.

For all install paths, see Installation.

11. Quick reference

CommandUse
estacodaStart an interactive session, or open setup if setup is incomplete
estacoda setupRun reviewed onboarding, setup, or repair
estacoda verifyCheck setup readiness
estacoda model statusInspect active provider/model configuration
estacoda doctorHealth report and required fixes
estacoda doctor --liveInclude live provider checks
estacoda --helpShow CLI help

What next