Nomos is a free, open source AI digital clone that learns who you are, acts on your behalf, remembers everything, and represents you across every platform. It deploys to iMessage, Slack, Discord, Telegram, and WhatsApp. Self-hosted on your own infrastructure with 60+ built-in skills and MIT licensed.

Is Nomos free to use?

Yes, Nomos is completely free and open source under the MIT license. You self-host it on your own infrastructure with full control over your data. Bring any LLM provider — Anthropic Claude, Google Vertex AI, OpenRouter, or run locally with Ollama. You only pay for the LLM API usage from your chosen provider.

What platforms does Nomos support?

Nomos deploys to iMessage, Slack, Discord, Telegram, WhatsApp, and Web/gRPC. With Slack User Mode, it sends messages as you — not as a bot — so teammates interact with your digital clone as if it were you. Setup takes minutes with the built-in web dashboard.

How does Nomos learn who I am?

Nomos uses adaptive memory powered by pgvector to extract your preferences, decisions, and communication style from every conversation. It builds a persistent model of you that deepens over time, with total recall across all channels and sessions. It learns how you write, what you care about, and how you make decisions.

How is Nomos different from other AI agents?

Most AI agents are chatbots or coding assistants that forget you between sessions. Nomos is a digital clone — it learns your identity, represents you authentically, and acts as you across platforms. Key differentiators include Slack User Mode (sends as you, not a bot), persistent adaptive memory, multi-provider support (Claude, Ollama, OpenRouter, Vertex AI), multi-agent team orchestration, and AES-256 encryption. It is self-hosted and MIT licensed.

What LLM providers does Nomos support?

Nomos supports multiple LLM providers including Anthropic Claude, Google Vertex AI, OpenRouter (access to hundreds of models), and Ollama for fully local inference. Smart model routing automatically directs simple queries to fast, cheap models and complex tasks to more capable ones, reducing costs by 5-10x.

Is my data safe with Nomos?

Nomos is fully self-hosted — your data never leaves your infrastructure. All memory and conversation data is encrypted at rest with AES-256-GCM. You control which LLM providers your data is sent to, and you can run fully local with Ollama for zero external data exposure.

Open Source · MIT License

Your AI Digital Clone

Ingests your communication history. Learns your voice. Builds a knowledge base about your world. Acts on your behalf across iMessage, Slack, Discord, Email, and more. Multi-provider, self-hosted, and fully yours.

Get Started View on GitHub


$ npx @project-nomos/nomos chat

Everything Your Clone Needs

History ingestion, voice modeling, knowledge compilation, identity graph, personality training, agent-to-agent trust, plugins, multi-provider routing, 65+ skills, and encryption — all built in, all open source.

Ingests Your History

Import years of Slack, Gmail, iMessage, and WhatsApp messages. Your clone starts with deep context from day one — not a blank slate. Continuous delta sync keeps it current.

Learns Your Voice

Per-contact style model analyzes formality, length, emoji, greetings, and sign-offs. Drafts messages in your authentic voice — different tone for your manager vs. your friends.

Compiles Knowledge

A Karpathy-style wiki transforms raw messages into structured articles about your contacts, projects, and topics. The clone reads compiled knowledge first, RAG second.

Unified Identity Graph

Links contacts across Slack, email, iMessage, Discord, and WhatsApp. One person, one profile, with per-contact autonomy: auto-send, draft, or silent.

Agent-to-Agent Trust

CATE protocol enables secure communication with other agents via DID-based identity, verifiable credentials, policy enforcement, and anti-spam stamps.

Acts As You

Drafts emails, manages your calendar, preps meeting briefs, tracks commitments, follows up. Per-contact autonomy controls what gets sent automatically.

Thinks Like You

Calibration interviews, decision pattern recognition, value hierarchy, theory of mind, adversarial twin testing, and personality DNA export. 10 training systems that teach the clone how you actually think and decide.

Portable Personality

Export your personality as a ~2000-token DNA document -- decision patterns, values, style genome, and exemplar fingerprints. Import into a fresh instance for instant cold-start.

Total Recall

Every conversation auto-indexed into pgvector. Ask about last week's discussion across any channel or session — it knows.

Represents You Everywhere

iMessage, Slack, Discord, Telegram, WhatsApp, Email, and web. Your clone is reachable from wherever you work.

Email Channel

IMAP IDLE for real-time inbox monitoring, SMTP for sending. Thread tracking, draft-and-approve for replies — like Slack User Mode but for email.

Passive Observation

Silently monitor Slack channels to learn communication patterns. Messages are indexed and analyzed, but the agent stays quiet — learning without interrupting.

Proactive Agency

Tracks commitments from conversations, generates pre-meeting briefs with attendee context, and triages unread across channels by sender importance.

Multi-Provider

Anthropic, Vertex AI, OpenRouter, Ollama, or any custom endpoint. Cloud models, local models, or mix and match.

Smart Model Routing

Route by complexity across any provider — simple queries to fast models, complex to capable ones. Cut costs 5-10x automatically.

Multi-Agent Teams

Coordinator + parallel workers. Hand off complex tasks and get synthesized results from multiple agents.

Image & Video Generation

Gemini image + Veo video generation built in. Just ask conversationally — no configuration needed.

Web Dashboard

Next.js settings UI with setup wizard. Configure integrations, models, and permissions — no YAML editing.

65+ Skills

Gmail, Calendar, Drive, Docs, Sheets, document gen, browser automation, and more. Three-tier loading: bundled, personal, project.

Plugin Ecosystem

Browse, install, and load plugins from the Claude Code ecosystem. 21 plugins (14 first-party + 7 community) available out of the box. Extends every SDK call -- CLI, daemon, and team workers.

Encrypted at Rest

AES-256-GCM for all API keys and integration tokens. Your clone stays on your infrastructure.

Self-Improvement

Nomos can analyze its own codebase, implement fixes, and open pull requests to itself — autonomously.

CLI / npx Interface

Run Nomos from the terminal with a single command. npx @project-nomos/nomos chat — no install required.

Browser Automation

Playwright-based interactive browser with persistent sessions. Navigate, click, type, screenshot — for login flows and scraping.

Task Orchestration

Task lifecycle tracking with dependency graphs, auto-unblock on completion, cycle detection, and abort-based cancellation.

LSP Code Intelligence

Go-to-definition, find-references, hover, and document symbols via TypeScript language server — built into the agent.

Sleep & Self-Resume

Agents pause and wake with a prompt. Poll deployments, wait for async operations, or monitor within a session.

Scheduled Tasks

Run agent prompts on a schedule — intervals, cron expressions, or one-time. Results delivered to any connected channel.

Proactive Messaging

The agent reaches out without being asked — urgent alerts, monitoring results, or scheduled task output to any channel.

Cost Tracking

Per-model pricing tiers, session cost summaries, and usage breakdown. Track spend across Haiku, Sonnet, and Opus in CLI and web dashboard.

Event Hooks

Extend the agent with command, HTTP, or prompt hooks. Trigger on tool use, session lifecycle, or compaction — exit code 2 blocks execution.

Bash Safety Analysis

Lightweight command analysis detects destructive operations, elevated privileges, and dangerous git commands before execution.

Magic Docs

Markdown files with a magic marker auto-update in the background when stale. Self-maintaining documentation powered by forked agents.

Context Visualization

See exactly how your context window is used — system prompt, conversation, tools, memory, skills — with color-coded bar chart and capacity warnings.

Auto-Dream Memory

Background memory consolidation triggered by time and turn count gates. Four-phase cleanup keeps your knowledge base fresh and accurate.

gRPC & WebSocket API

Build custom clients on top of your clone. Bidirectional streaming via gRPC, plus legacy WebSocket support — mobile, web, or CLI.

Up and Running in Minutes

From install to deployed digital clone in five steps.

Install & Connect

One command to start. The setup wizard connects your database, API key, and communication channels.

$ npx @project-nomos/nomos chat

Ingest Your History

Import years of Slack, Gmail, iMessage, and WhatsApp messages. Your clone starts with deep context from day one.

$ nomos ingest imessage --since 2024-01-01

It Learns Your Voice

Analyzes your writing style per contact, compiles a knowledge wiki about your world, and builds a unified identity graph across all platforms.

Deploy Everywhere

Run the daemon to connect Slack, Discord, Telegram, Email, and more. Your clone is always on, always reachable.

$ nomos daemon start

It Acts As You

Drafts in your voice, tracks commitments, preps meeting briefs, triages across channels — or hands off complex tasks to multi-agent teams.

Smart Model Selection

The Right Model for Every Message

Nomos analyzes each message and routes it to the best model based on complexity. Simple lookups go to fast models. Deep reasoning goes to premium ones. You set the providers — Nomos handles the rest.

Complexity-Based Routing

Simple questions go to fast, cheap models. Complex reasoning gets routed to premium models. Automatic — no rules to configure.

Cut Costs Automatically

Use powerful models only when needed. Route 80% of queries to lightweight models and save without sacrificing quality.

5 Providers, Any Model

Anthropic, Vertex AI, OpenRouter, Ollama, or custom endpoints. Mix cloud and local models in a single agent.


You: What time is it in Tokyo?
→ routed to gemini-flash · simple lookup
You: Analyze our Q4 revenue trends and suggest strategies
→ routed to claude-sonnet · complex reasoning
You: Generate an image of a mountain sunset
→ routed to gemini-pro · image generation
✓ 3 queries · 2 providers · 67% cost saved vs. premium-only

Deploy to 7 Platforms

One agent runtime, thin adapters. Messages from all channels flow into the same brain with the same memory.

Slack

Available

Socket Mode + User Mode with draft approval

Discord

Available

Full bot integration with streaming responses

Available

Built on grammY with markdown rendering

Available

Via Baileys — no business API needed

iMessage

macOS

macOS native — reads from Messages.app

Email

Available

IMAP IDLE for real-time push, SMTP for sending

Web / gRPC

Available

Connect any client via gRPC or WebSocket


Settings Dashboard
Primary Modelclaude-sonnet-4
Fast Modelgemini-flash
Smart RoutingEnabled
SlackConnected
DiscordConnected
TelegramPending
Skills Enabled28 / 35
EncryptionAES-256-GCM

Web Dashboard

Configure Everything from the Browser

No YAML files. No environment variables. The built-in Next.js dashboard gives you full control over models, integrations, skills, and permissions — with a setup wizard that gets you running in minutes.

Model & Provider Config

Switch models, set routing rules, and configure API keys — all from the UI.

Channel Integrations

Connect Slack, Discord, Telegram, and WhatsApp with guided setup. No config files.

Permissions & Safety

Control which skills are enabled, set content filters, and manage user access.

Personality & Behavior

Customize the agent's tone, system prompts, and response style per channel.

Setup Wizard

First-run wizard walks you through database, API keys, and channel setup step by step.

An Agent That Evolves Itself

Nomos has a built-in self-improvement skill. It can analyze its own codebase, implement changes, and open pull requests — all autonomously.

Clones a fresh copy — never modifies the running instance

Runs all checks: format, typecheck, lint, and tests

Opens a PR for your review with a clear summary

Just say "improve yourself" or "add tests for the chunker"


You:
Add edge case tests for the memory chunker
Nomos:
I'll analyze the chunker, write tests, and open a PR.
✓ Cloned repo to /tmp/nomos
✓ Analyzed src/memory/chunker.ts
✓ Added 6 edge case tests
✓ All 199 tests passing
✓ PR opened: #42

Advanced Agent Capabilities

Personality training, adversarial testing, agent-to-agent trust, plugins, browser automation, task orchestration, code intelligence, and autonomous operations.

Historical Data Ingestion

Import years of Slack, Gmail, iMessage, and WhatsApp history. Deduplication, chunking, embedding, and continuous delta sync — your clone starts with deep context from day one.

nomos ingestnomos ingest status

Communication Style Model

Per-contact voice analysis extracts formality, length, emoji usage, greetings, and sign-offs from your sent messages. The clone drafts in your authentic voice — different tone for every relationship.

Knowledge Wiki

A Karpathy-style compiled knowledge base. An LLM periodically transforms raw messages into structured articles about contacts, projects, and topics. Wiki-first, RAG-fallback retrieval.

wiki_search

Cross-Channel Identity

Unified contacts linking Slack, email, phone, Discord, and more into a single profile. Auto-linking heuristics, per-contact autonomy levels (auto/draft/silent), and privacy controls.

nomos contacts

CATE Protocol

Secure agent-to-agent communication via Consumer Agent Trust Envelopes. DID-based identity, verifiable credentials, policy engine, and anti-spam stamps (PoW or micropayment).

Passive Observation

Silently observe Slack channels to learn communication patterns without triggering responses. Messages are indexed and analyzed but the agent stays quiet.

Proactive Agency

Commitment tracking, pre-meeting briefs with attendee lookup, and cross-channel priority triage. Your clone anticipates what you need before you ask.

Calibration Interviews

Scenario-based interviews across 10 domains (tech decisions, communication, conflict, prioritization, leadership, quality, collaboration, risk, creativity, time management). Socratic follow-ups extract decision patterns at 0.85 confidence.

/calibrate/calibrate status

Decision & Value Extraction

Automatic extraction of decision heuristics and core values from every conversation via lightweight LLM call. Confidence-weighted merging -- repeated confirmations boost, contradictions reduce. Context preserved per domain.

user_model_recall

Adversarial Twin Testing

GAN-style clone fidelity testing. Pulls real user messages, generates clone responses blind, then a discriminator identifies which is real. Style corrections feed back into the user model at 0.8 confidence.

/twin-test

Theory of Mind & Personas

Real-time rule-based model of the user's mental state (focus, emotion, cognitive load, urgency, energy) injected per message. Contextual personas switch tone, formality, and emoji usage based on platform, time, keywords, and contacts.

Reflection Cycles

The clone articulates its current model of you -- decision-making style, core values, communication preferences. Makes scenario-specific predictions with confidence scores. Identifies blind spots and accepts corrections.

/reflect

Personality DNA

Compiles the full personality model into a portable ~2000-token JSON document. Top 10 decision patterns, top 5 values, style genome, and exemplar fingerprints. Import into a fresh instance for instant cold-start at 0.7 confidence.

/dna export/dna import/dna diff

Shadow Mode

Opt-in passive observer that records tool usage sequences, corrections, file access patterns, and response cadence without storing message content. After 20+ observations, distills behavioral patterns into the user model via Haiku.

Plugin Ecosystem

Browse, install, and load plugins from the Claude Code marketplace. 21 plugins (14 first-party + 7 community) fetched at install time. Passed as SdkPluginConfig[] to every SDK query() call across CLI, daemon, and team workers.

nomos plugin listnomos plugin install

Interactive Browser

Playwright-based browser automation with persistent sessions. Navigate, click, type, screenshot, and execute JavaScript across multi-step workflows.

browser_navigatebrowser_clickbrowser_typebrowser_screenshot

Task Dependency Graph

Every operation tracked as a task with lifecycle states. Declare dependencies between tasks — they auto-start when blockers complete, with cycle detection.

task_statustask_kill

LSP Code Intelligence

TypeScript language server integration. Go-to-definition, find-references, hover for type info, and document symbols — without leaving the conversation.

lsp_go_to_definitionlsp_find_referenceslsp_hover

Plan Mode

For complex tasks, the agent proposes a structured plan with steps, affected files, and risk levels. Review and approve before any changes are made.

propose_plan

Sleep & Self-Resume

Agents pause execution for up to an hour and resume with a wake-up prompt. Perfect for polling deployments or waiting for async operations.

agent_sleep

Proactive Messaging

Send messages to your channels unprompted — urgent alerts, monitoring results, or scheduled task output. Routes through the daemon's channel manager.

proactive_send

Inter-Agent Messaging

Workers communicate via an in-memory message bus with priority levels during team execution. Share results or report blockers in real time.

send_worker_messagecheck_worker_messages

Verification Agent

After team workers finish, a read-only adversarial agent runs tests, build, and lint. Reports PASS, FAIL, or PARTIAL with details.

Memory Consolidation

Four-phase cleanup: prune stale chunks, merge near-duplicates by vector similarity, LLM-powered review via Haiku, and confidence decay.

memory_consolidate

Worktree Isolation

Team workers operate in isolated git worktrees to avoid conflicts. Auto-cleanup if no changes were made; preserved for review if changes exist.

Context Visualization

See exactly how your context window is being used — system prompt, conversation, tools, memory, and skills. Bar chart in CLI and web dashboard with capacity warnings.

Hook System

Event-driven extensibility via hooks.json. Command, HTTP webhook, and prompt hooks fire on tool use, session lifecycle, or compaction. Pattern matching with glob support.

Auto-Dream Consolidation

Background memory consolidation triggered by time and turn count gates. Four phases: Orient, Gather, Consolidate, Prune. Lock-file coordination prevents concurrent runs.

Adaptive Retry

Exponential backoff with jitter, retry-after header parsing, and 429/529 detection. Persistent mode for daemon retries indefinitely with abort signal support.

Email Integration

Full email channel with IMAP IDLE for real-time inbox push and SMTP for sending. Thread tracking via In-Reply-To headers, HTML-to-text for LLM processing, draft-and-approve workflow.

65+ Skills, Three-Tier Loading

Bundled skills get you started. Add personal skills at ~/.nomos/skills/ or project-specific ones at ./skills/. Create your own in minutes.

Productivity

Google CalendarGoogle DocsGoogle DriveGmailGoogle SheetsGoogle SlidesApple NotesApple RemindersDoc Co-authoringWeather

Communication

SlackDiscordTelegramWhatsAppDigital MarketingInternal Comms

Development

GitHubMCP BuilderWebapp TestingFrontend DesignSelf-ImproveSkill Creator

Creative

Image GenerationVideo GenerationAlgorithmic ArtBrand GuidelinesTheme FactoryCanvas Design

Documents

PDFDOCXPPTXXLSXWeb Artifacts Builder

Identity Training

Calibration InterviewsReflection CyclesTwin TestPersonality DNA

Operations

Browser AutomationScheduled TasksCost TrackingBash SafetyMagic DocsAuto-DreamPlugin Manager

Ready to Build Your Clone?

Import your history, learn your voice, deploy everywhere. Open source, self-hosted, MIT licensed — your AI digital clone, running on your infrastructure.

Get Started Star on GitHub