Skip to content

Your Private AI in 5 Minutes

Get your own private AI assistant and start chatting with it on Signal — in under 5 minutes. Free plan, no credit card required.

Alpha Access Required

Lucid is in private alpha. Request access before proceeding.

Time to complete: ~5 minutes

What You'll Get

  • A private AI assistant powered by Llama 3.1 8B, running in encrypted hardware
  • Signal integration — chat with your AI from your phone
  • Hardware-verified security — TEE attestation on every message
  • AI Passport — cryptographic proof of privacy and compliance
  • Free tier — 50 messages/day, no credit card required

Step 1: Install and Sign Up

pip install lucid-clilucid loginEmail: [email protected]
Password: ********

Logged in as [email protected]
Plan: Free (Individual)

The free plan includes:

Feature Included
Private agents 1
Model Llama 3.1 8B
Messages/day 50
Auditors Injection, Toxicity
Hardware security TEE (AMD SEV-SNP)

Step 2: Deploy Your Agent

lucid apply --model llama-3.1-8b --profile chat[*] Creating your private agent...
[+] Agent created: agent-your-abc123

Status: running
Model: meta-llama/Llama-3.1-8B-Instruct
Auditors: injection, toxicity
Runtime: nanobot
Hardware: AMD SEV-SNP (TEE attested)

[+] Agent ready! Connect a messaging channel to start chatting.

Your agent is now running in a hardware-secured enclave. Every message is audited — injection detection and toxicity filtering are active by default.

Step 3: Connect Signal

Link your agent to Signal so you can chat from your phone:

lucid agent connect signal[*] Generating Signal link...

Scan this QR code with Signal on your phone:

██████████████████████████
██ ██
██ ▄▄▄▄▄ ██ ▄▄▄▄▄ ██
██ █ █ ██ █ █ ██
██ ▀▀▀▀▀ ██ ▀▀▀▀▀ ██
██ ██
██████████████████████████

Or enter this code in Signal:
lucid-agent-abc123

[+] Waiting for Signal connection...
[+] Connected! Send a message to start chatting.

Open Signal on your phone, scan the QR code, and you're connected.

Step 4: Start Chatting

Send a message to your agent on Signal. Try these:

You say What happens
"Hey, what can you do?" Agent introduces its capabilities
"Summarize the latest AI news" Agent searches and summarizes
"Remind me to call the dentist tomorrow" Agent creates a reminder
"What's 15% tip on $84.50?" Agent does quick math

Every message passes through the auditor chain. Injection attempts and toxic content are automatically blocked before the model sees them.

Step 5: View Your AI Passport (Optional)

Your agent has a live AI Passport — cryptographic proof of its security and compliance:

lucid passport showPassport for: agent-your-abc123

Status: VERIFIED
Hardware: AMD SEV-SNP (TEE attested)
Model: Llama 3.1 8B — runs in encrypted memory
Auditors Active: injection (pass), toxicity (pass)
Messages Today: 12 / 50
Data Residency: us-east-1
Last Verified: 2026-03-07T14:30:00Z

Or view it in the Observer dashboard:

lucid open[+] Opening Observer dashboard...

The Observer shows your agent's real-time status, audit trail, and passport in a web UI.

Step 6: View Policies

See what rules your agent follows:

lucid agent policiesPOLICY STATUS DESCRIPTION
injection-guard active Blocks prompt injection attacks
toxicity-filter active Filters harmful or toxic content
pii-protection active Redacts personal data before model sees it
rate-limit active 50 messages/day on free plan

These are enforced automatically. View detailed policy rules in Observer under Settings > Policies.

Other Messaging Channels

Signal not your thing? Connect other channels the same way:

lucid agent connect telegramlucid agent connect whatsapp

Each channel gets the same auditor enforcement, the same passport, the same privacy guarantees.

Upgrade: Bring Your Own API Key

Want a more powerful model? Bring your own Anthropic (or OpenAI / Groq) API key — your agent upgrades instantly.

Add Your Key

lucid registry add-key anthropicPaste your Anthropic API key: sk-ant-••••••••
[+] Key registered (masked: sk-ant-****7f2a)

⚠ Privacy notice: prompts and responses will be sent to
Anthropic's API. Auditors still run inside the enclave.

Redeploy with Claude

lucid apply --model claude-sonnet-4-6 --provider anthropic[*] Redeploying agent-your-abc123...
[+] Agent updated:

Model: claude-sonnet-4-6 (via Anthropic API)
Auditors: injection, toxicity (in-TEE)
Hardware: AMD SEV-SNP (auditor chain attested)
Inference: external/anthropic

[+] Agent ready — same Signal connection, upgraded model.

What Changes in the Passport

Your AI Passport updates to reflect the new trust boundary:

Field Private (Llama 8B) External (Claude)
Inference provider private external/anthropic
Inference confidentiality enclave provider_visible
Data boundary enclave_only enclave_and_external_api
TEE mode in-TEE in-TEE (auditors only)
Auditor attestation Hardware-attested Hardware-attested

The passport is honest — it tells you exactly what's private and what isn't.

Switch Back Anytime

lucid apply --model llama-3.1-8b[+] Agent reverted to fully private mode.

No data leaves the enclave when using the built-in model.

Upgrade for More

Feature Free Pro ($15/mo) Max ($49/mo)
Messages/day 50 300 Unlimited
Models 8B 8B, 70B, 72B All
Channels 1 Unlimited Unlimited
Custom policies — Yes Yes
API access — — Yes

Upgrade anytime:

lucid billing upgrade pro

Clean Up

Stop your agent:

lucid stop agent-your-abc123

Or delete it entirely:

lucid agent delete agent-your-abc123

Next Steps

Summary

Component Value
Time to complete ~5 minutes
Plan Free (no credit card)
Model Llama 3.1 8B Instruct
Runtime Nanobot
Auditors Injection, Toxicity
Channel Signal (or Telegram, WhatsApp)
Messages/day 50
Hardware TEE-attested (AMD SEV-SNP)