Integration Guide

This guide covers how to connect your local development tools to deployed Lucid agents, enabling observability, TEE verification, and secure model access from your terminal.

Alpha Access Required

Lucid is in private alpha. Some integrations described here may not yet be available. Request access to get started.

Looking for deployment instructions?

See the Deployment Guide to deploy agents first, then return here to connect your tools.

Overview

Lucid supports integrating with popular development tools so you can:

Route requests through your deployed agent's auditor chain
Verify TEE attestation from your local environment
Get observability into your coding sessions via AI Passport

Tool	Type	Best For
OpenCode	Terminal AI	Fast, lightweight coding assistance
OpenClaw	Desktop Agent	Local terminal and OS automation
Aider	AI Pair Programming	Git-aware code editing
Claude Code	CLI	Anthropic's official Claude CLI

OpenCode

OpenCode is a Go-based terminal AI assistant optimized for speed and low resource usage. Connect it to your Lucid agent for audited, TEE-verified coding assistance.

Installation

go install github.com/opencode-ai/opencode@latestgo: downloading github.com/opencode-ai/opencode v1.2.0

Or via Homebrew:

brew install opencode-ai/tap/opencode

Connect to Your Agent

Get your agent's endpoint from the Observer UI or CLI:

lucid status my-agentNAME STATUS ENDPOINT
my-agent running https://my-agent.env-abc123.lucid.ai
lucid credentials my-agentAPI Key: sk-lucid-xxxxxxxxxxxx

Configure OpenCode to use your agent:

opencode config set --endpoint https://my-agent.env-abc123.lucid.aiEndpoint configured: https://my-agent.env-abc123.lucid.ai
opencode config set --api-key sk-lucid-xxxxxxxxxxxxAPI key configured (hidden)

Verify Connection

opencode ping[+] Connected to Lucid agent
Endpoint: https://my-agent.env-abc123.lucid.ai
Model: Qwen/Qwen2.5-Coder-32B-Instruct
TEE: AMD SEV-SNP (verified)
Auditors: injection, eval

Usage

Once configured, use OpenCode as normal. All requests route through your Lucid agent:

opencode "refactor this function to use async/await"Analyzing codebase...
[Lucid] Request audited: injection=pass, eval=pass

Here's the refactored version:
...

Configuration Reference

Setting	Command	Description
Endpoint	`opencode config set --endpoint <url>`	Lucid agent URL
API Key	`opencode config set --api-key <key>`	Agent API key
Model	`opencode config set --model <id>`	Override model (optional)
Verify TEE	`opencode config set --verify-tee true`	Require TEE attestation

OpenClaw

OpenClaw is a desktop agent for local terminal and OS automation. It runs locally but connects to your Lucid agent for model inference, giving you TEE-verified AI control over your development environment.

Installation

Coming Soon

OpenClaw installation packages are being finalized for alpha participants. Request access for early availability.

curl -fsSL https://get.lucidcomputing.ai/openclaw | sh

Connect to Your Agent

openclaw initWelcome to OpenClaw!

Enter your Lucid agent endpoint: https://my-agent.env-abc123.lucid.ai
Enter your API key: sk-lucid-xxxxxxxxxxxx

[+] Verifying TEE attestation...
[+] Agent verified: AMD SEV-SNP
[+] Configuration saved to ~/.openclaw/config.yaml

Verify Setup

openclaw doctorOpenClaw Health Check
=====================

[+] Configuration: OK
Endpoint: https://my-agent.env-abc123.lucid.ai
API Key: sk-lucid-****xxxx (configured)

[+] Connectivity: OK
Latency: 45ms
Model: moonshotai/Kimi-K2.5

[+] TEE Verification: OK
Type: AMD SEV-SNP
Attestation: Valid
Measurement: sha256:abc123...

[+] Permissions: OK
Shell access: Enabled
File system: ~/projects (sandboxed)

All checks passed!

Usage Examples

Run shell commands with AI assistance:

openclaw "find all TODO comments in this project"[Lucid] Request audited: injection=pass

Searching for TODO comments...

Found 12 TODOs:
src/api/routes.ts:45 - TODO: Add rate limiting
src/db/migrations.ts:12 - TODO: Add index on user_id
...

Automate file operations:

openclaw "rename all .jsx files to .tsx and update imports"[Lucid] Request audited: injection=pass

Planning changes:
1. Rename 8 files from .jsx to .tsx
2. Update 23 import statements

Proceed? [y/N]: y

[+] Renamed: src/components/Button.jsx → Button.tsx
[+] Renamed: src/components/Modal.jsx → Modal.tsx
...
[+] Updated 23 imports across 12 files

Done! All changes committed to git.

Interactive terminal session:

openclaw shellOpenClaw Shell (type 'exit' to quit)
Connected to: my-agent (Kimi K2.5)

openclaw> deploy this app to staging
[Lucid] Request audited: injection=pass

I'll help deploy to staging. Let me check your configuration...

Found: docker-compose.staging.yml
Running: docker-compose -f docker-compose.staging.yml up -d

Deployment complete! Services running:
- api: healthy (port 3000)
- worker: healthy
- redis: healthy

Security Model

OpenClaw operates with a sandboxed permission model:

Permission	Default	Description
Shell commands	Prompt	Ask before executing commands
File read	Allowed	Read files in project directory
File write	Prompt	Ask before modifying files
Network	Blocked	No direct network access
System	Blocked	No system-level changes

Configure permissions in ~/.openclaw/config.yaml:

permissions:
  shell:
    mode: prompt        # prompt, allow, deny
    allowlist:
      - git
      - npm
      - docker
  filesystem:
    read: allow
    write: prompt
    root: ~/projects    # Sandbox root
  network: deny
  system: deny

Configuration Reference

Setting	Description
`endpoint`	Lucid agent URL
`api_key`	Agent API key
`verify_tee`	Require TEE attestation (default: true)
`model`	Override model selection
`permissions`	Security sandbox settings
`shell`	Default shell (bash, zsh, fish)

Aider

Aider is an AI pair programming tool that works with your git repository. Connect it to your Lucid agent for audited code editing.

Installation

pip install aider-chatSuccessfully installed aider-chat-0.50.0

Connect to Your Agent

Set environment variables to route through Lucid:

export OPENAI_API_BASE=https://my-agent.env-abc123.lucid.ai/v1
export OPENAI_API_KEY=sk-lucid-xxxxxxxxxxxx

Or create a .env file in your project:

OPENAI_API_BASE=https://my-agent.env-abc123.lucid.ai/v1
OPENAI_API_KEY=sk-lucid-xxxxxxxxxxxx

Usage

aiderAider v0.50.0
Model: Qwen/Qwen2.5-Coder-32B-Instruct (via Lucid)

> Add input validation to the user registration endpoint

I'll add validation to the registration endpoint...

Claude Code

Claude Code is Anthropic's official CLI for Claude. Connect it to Lucid for observability and passport verification.

Installation

npm install -g @anthropic-ai/claude-codeadded 1 package in 2s

Connect to Lucid Observer

Configure Claude Code to report telemetry to your Lucid Observer:

claude config set observer_endpoint https://observer.your-domain.comObserver endpoint configured
claude config set observer_key sk-observer-xxxxxxxxxxxxObserver key configured

Verify Setup

claude doctorClaude Code Health Check
========================

[+] Anthropic API: OK
[+] Observer: OK
Endpoint: https://observer.your-domain.com
Telemetry: Enabled

All checks passed!

Usage

With Observer configured, your Claude Code sessions appear in the Lucid Observer dashboard with full AI Passport tracking.

Lucid Browser Extension

For browser-based AI agents, install the Lucid Browser Extension to verify AI Passport authenticity.

Installation

Coming Soon

The browser extension is under development. Request access for early availability.

Click the Lucid extension icon
Go to Settings → Agent Connection
Configure your agent URL and API key

Features

Feature	Description
Passport Verification	Validates AI agent passports on visited pages
TEE Badges	Shows TEE status for verified agents
Action Approval	Human-in-the-loop for Browser Use agents

Troubleshooting

Connection Issues

Error: "Connection refused"

opencode pingError: Connection refused to https://my-agent.env-abc123.lucid.ai
lucid status my-agentNAME STATUS ENDPOINT
my-agent stopped -
lucid start my-agentAgent 'my-agent' started.

Error: "TEE verification failed"

Ensure your agent is running in a TEE-enabled environment:

lucid verify endpoint https://my-agent.env-abc123.lucid.ai[!] TEE verification failed: No attestation report

Check that your agent is deployed with TEE enabled:
lucid apply -f my-env.yaml # Ensure teeMode is set

Authentication Issues

Error: "Invalid API key"

Regenerate your API key:

lucid credentials my-agent --regenerateNew API Key: sk-lucid-yyyyyyyyyyyy

Update your tool configuration with the new key.

Next Steps

Deployment Guide - Deploy your first agent
AI Passports - Understand passport verification
Auditor Development - Create custom auditors