@greatlyrecommended/tigerpaw

v2026.4.3

Published

5 days ago

Local-first AI gateway — everything OpenClaw does, plus trading across 9 exchanges, risk controls, a React 19 dashboard, MCP support, and an AI assistant

Features

For everyone (even if you don't trade)

20+ Agent Channels (inherited from OpenClaw) -- Telegram · Discord · Slack · Signal · iMessage · WhatsApp · Matrix · MS Teams · IRC · Line · Nostr · Google Chat · Mattermost · Twitch · Feishu · Zalo and more
React 19 Control Dashboard -- Modern UI replacing OpenClaw's Lit/Web Components -- real-time positions, P&L charts, TradingView embeds, order entry, risk management, and approval queue
Gateway Security Hardening -- CORS allowlisting, per-IP rate limiting, request size enforcement, credential rotation tracking
Zero-Config Start -- tigerpaw start creates config, starts the gateway, and opens the dashboard. One command.
Docker Multi-Arch Images -- amd64 + arm64 with rootless Podman/systemd support
Plugin Permission Manifests -- Declarative permission model for extensions (network, trading, filesystem, secrets) with security audit via tigerpaw doctor
Local-First by Default -- Gateway binds to localhost; API keys and data never leave your machine
AI Assistant (Jarvis) -- Task management, reminders, daily briefings, and knowledge retrieval
Message Hub -- Unified inbox across all agent channels with search, filtering, and date grouping
Visual Workflow Builder -- Drag-and-drop event-driven automation (trading events, cron schedules, message routing) with per-node input/output inspection, real-time execution monitoring, single-node testing, data pinning, and debug replay
Expression Engine & Data Mapping -- 35 built-in functions (string, array, math, date, logic), {{expression}} templates with autocomplete, and visual upstream node output mapping
Integration SDK -- Extensible framework for adding new integrations. 6 built-in providers (HTTP Request, Slack, Google Sheets, GitHub, OpenAI, Anthropic) with standardized auth, schema-driven UI, and rate limiting
MCP Protocol Support -- Connect external tool servers (stdio + SSE) and expose Tigerpaw tools to external AI agents
Local LLM Support -- Ollama and LM Studio with auto-detection, model management, and cloud fallback
i18n in 10 Languages -- English, Spanish, French, German, Japanese, Korean, Chinese (Simplified + Traditional), Portuguese, Arabic
Email, Calendar & Meeting Integrations -- Connect Gmail, Outlook, Google Calendar, Zoom, Google Meet, and Microsoft Teams via OAuth2. Jarvis can read emails, create events, and schedule meetings. Workflow engine includes send_email, create_calendar_event, and schedule_meeting action nodes.
Code Splitting -- 21 lazy-loaded routes with 354KB max chunk size for fast initial load

For traders

9 Trading Platforms -- Alpaca · Polymarket · Kalshi · Manifold · Coinbase · Interactive Brokers · Binance · Kraken · dYdX
Policy-Gated Trading -- Every order goes through a 12-step validation pipeline before execution
Risk Management -- Daily spend limits, position limits, drawdown protection, cooldowns, kill switch
Tamper-Evident Audit Log -- HMAC-SHA256 chain-linked JSONL logging for every trade decision
3 Approval Modes -- Auto, confirm (configurable timeout + deny/approve on timeout), or manual (configurable timeout)
3 Risk Tiers -- Conservative, moderate, aggressive presets
Real-Time Notifications -- In-app toast alerts for order approvals, denials, kill switch changes, and limit warnings
Trading Agent Commands -- 8 unified trading tools accessible from any agent channel -- portfolio summary, P&L, positions, kill switch, risk status
Remote Dashboard Access -- Access your dashboard from any device via Tailscale (end-to-end encrypted) or Cloudflare Tunnel (free HTTPS)
Per-Platform Risk Overrides -- Override any limit or approval mode per exchange
Daily P&L Charts + Position Heatmaps -- Visual portfolio analysis in the dashboard
Security Audit Dashboard -- Audit findings, credential rotation status, and extension permission checks
Close Position Dialog -- One-click exit with policy pre-check

Why Tigerpaw?

The Problem

70-80% of retail traders lose money. The most common causes aren't bad strategies -- they're behavioral: overtrading, ignoring stop losses, risking too much per trade, and emotional decisions after losses. Cloud-hosted trading bots make this worse by operating as black boxes with no guardrails -- one misconfigured algorithm can drain an account in minutes.

Meanwhile, cloud trading platforms introduce security risks that most traders don't think about:

API key exposure: Stolen credentials drove 22% of breaches in 2025. Cloud platforms store your exchange API keys on their servers -- one breach and every connected account is compromised.
No data sovereignty: Your trading history, positions, and strategies live on someone else's infrastructure.
Single point of failure: When a cloud provider goes down, you lose access to your positions with no manual override.

The Solution

Tigerpaw runs entirely on your machine. Your API keys never leave your computer. Your trading data stays on your disk. The gateway binds to localhost by default -- it's not even reachable from your local network, let alone the internet.

But local-first alone isn't enough. Tigerpaw adds institutional-grade risk controls that prevent the mistakes that wipe out retail accounts:

| Protection | What It Prevents | | -------------------------------- | ---------------------------------------------------------------------------- | | Daily spend cap | Can't accidentally deploy your entire portfolio in one day | | Per-trade size limit | No single trade can risk more than a configured % of your portfolio | | Position concentration limit | Can't go all-in on one asset | | Daily loss limit | Auto-activates kill switch when losses hit your threshold | | Drawdown protection | Stops trading when portfolio drops from its peak | | Cooldown timer | Prevents revenge trading after a loss | | Consecutive loss pause | Automatic pause after N losing trades in a row | | Kill switch | Instant halt -- manual or auto-triggered, blocks all new orders | | Approval modes | Require manual confirmation before every trade (or auto-approve with limits) |

These aren't optional add-ons -- they're built into the core. Every order goes through a 12-step validation pipeline before it can execute. There is no way to bypass it.

Why Tigerpaw instead of OpenClaw?

Tigerpaw is a strict superset of OpenClaw. Every feature OpenClaw has -- 40+ messaging channels, agent runtime, plugin system -- is included in Tigerpaw. On top of that, Tigerpaw adds:

A modern React 19 dashboard (OpenClaw uses older Lit/Web Components)
Gateway security hardening (CORS, rate limiting, request size enforcement, credential rotation)
Zero-config tigerpaw start experience
Docker multi-arch images (amd64 + arm64)
Declarative plugin permission manifests with security audit
AI assistant (Jarvis) with daily briefings
Message Hub for unified cross-channel messaging
Visual workflow builder for event-driven automation
MCP protocol support (client + server)
Local LLM support with Ollama/LM Studio auto-detection
i18n in 10 languages
Code splitting with 354KB max chunk for fast load times
And for traders: a full policy-gated trading engine across 9 exchanges

If you're choosing between them, Tigerpaw gives you everything OpenClaw has, plus significantly more.

Who Is This For?

Anyone using OpenClaw -- Tigerpaw is a drop-in upgrade with a better UI, stronger security, and zero-config start
Developers building AI agents or trading systems who need a multi-channel messaging gateway
Quantitative traders who want local execution with institutional-style risk management
Privacy-conscious users who don't want their API keys or data on third-party servers
AI agent builders -- MCP integration lets you wire Tigerpaw into any AI tool ecosystem

"The best risk management is the kind you can't turn off." -- Tigerpaw's kill switch auto-activates when limits are breached. You can't accidentally trade through a drawdown.

Support

If you find Tigerpaw useful, consider supporting development:

Screenshots

Install

Requires Node.js 22+.

# 1. Install globally via npm
npm install -g @greatlyrecommended/tigerpaw

# 2. Verify installation
tigerpaw --version

# 3. Start the gateway + dashboard
tigerpaw start

That's it. tigerpaw start creates a config with safe defaults (paper mode, localhost-only, conservative risk tier), starts the gateway, and opens the dashboard at http://localhost:18789.

From Source

git clone https://github.com/varunrazdan/tigerpaw.git
cd tigerpaw
pnpm install
pnpm build
node tigerpaw.mjs start

Docker

docker run -d --name tigerpaw \
  -p 18789:18789 \
  -e TIGERPAW_GATEWAY_TOKEN=your-secret-token \
  ghcr.io/varunrazdan/tigerpaw:latest \
  tigerpaw start --bind lan

Update to Latest

npm update -g @greatlyrecommended/tigerpaw

Quick Start

After installation, the typical workflow is:

# Start gateway + open dashboard
tigerpaw start

# Or with manual setup:
tigerpaw setup                     # Create config + workspace
tigerpaw gateway run --open        # Start gateway + open browser
tigerpaw doctor                    # Check system health + security audit

First Run

On first launch, an onboarding wizard guides you through:

AI provider — connect OpenAI, Anthropic, Ollama, or another LLM
Messaging — connect Discord, Telegram, Slack, Signal, or WhatsApp
Trading — connect Alpaca, Coinbase, Binance, or other exchanges
Demo mode — start with sample data to explore the dashboard

Configuration

Config lives at ~/.tigerpaw/tigerpaw.json.

tigerpaw config set gateway.port 18789
tigerpaw config get

Environment Variables

| Variable | Purpose | | ------------------------ | ------------------------------------------------- | | TIGERPAW_STATE_DIR | Override state directory (default: ~/.tigerpaw) | | TIGERPAW_GATEWAY_PORT | Gateway port override | | TIGERPAW_GATEWAY_TOKEN | Gateway auth token |

Credentials use SecretRef syntax -- never hardcode API keys:

{ "apiKey": "${MY_ENV_VAR}" }

Trading Setup

Add a trading block to your config:

{
  "trading": {
    "enabled": true,
    "mode": "paper",
    "policy": {
      "tier": "conservative",
      "approvalMode": "confirm",
      "confirm": { "timeoutMs": 30000, "showNotification": true, "timeoutAction": "deny" },
      "manual": { "timeoutMs": 300000, "timeoutAction": "deny" },
      "limits": {
        "maxDailySpendUsd": 100,
        "maxSingleTradeUsd": 25,
        "maxTradesPerDay": 10,
        "maxOpenPositions": 3,
        "maxRiskPerTradePercent": 1,
        "maxSinglePositionPercent": 5,
        "dailyLossLimitPercent": 3,
        "maxPortfolioDrawdownPercent": 10,
        "cooldownBetweenTradesMs": 60000,
        "consecutiveLossPause": 3
      },
      "perExtension": {
        "alpaca": { "maxSingleTradeUsd": 50, "approvalMode": "manual" },
        "manifold": { "approvalMode": "auto" }
      }
    },
    "auditLog": {
      "maxFileSizeMb": 50,
      "rotateCount": 5
    }
  }
}

Trading Config Reference

| Field | ------------------------------------ | trading.enabled | trading.mode | policy.tier | policy.approvalMode | policy.confirm.timeoutMs | policy.confirm.timeoutAction | policy.confirm.showNotification | policy.manual.timeoutMs | policy.manual.timeoutAction | limits.maxDailySpendUsd | limits.maxSingleTradeUsd | limits.maxTradesPerDay | limits.maxOpenPositions | limits.maxRiskPerTradePercent | limits.maxSinglePositionPercent | limits.dailyLossLimitPercent | limits.maxPortfolioDrawdownPercent | limits.cooldownBetweenTradesMs | limits.consecutiveLossPause | policy.perExtension.<name> | auditLog.maxFileSizeMb | auditLog.rotateCount | Type | Default | Description | | ------------------------------------------------------------- | ---------------- | ----------------------------------------------- | | boolean | false | Enable the trading subsystem | | "paper" / "live" | "paper" | Paper simulates; live uses real money | | "conservative" / "moderate" / "aggressive" / "custom" | "conservative" | Risk preset (custom = manual limits) | | "auto" / "confirm" / "manual" | Varies by tier | How orders are approved | | number | 15000 | Confirm mode timeout (ms) | | "approve" / "deny" | "deny" | What happens when confirm times out | | boolean | true | Show UI notification for confirm requests | | number | 300000 | Manual approval timeout (ms) | | "approve" / "deny" | "deny" | What happens when manual approval times out | | number | Tier-dependent | Max cumulative daily notional spend (USD) | | number | Tier-dependent | Max single order size (USD) | | number | Tier-dependent | Max trades per calendar day (UTC) | | number | Tier-dependent | Max concurrent open positions | | number | Tier-dependent | Max % of portfolio risked per trade | | number | Tier-dependent | Max % of portfolio in one asset | | number | Tier-dependent | Daily loss trigger (% of portfolio) | | number | Tier-dependent | Drawdown from high-water mark trigger (%) | | number | Tier-dependent | Min time between trades (ms) | | number | Tier-dependent | Consecutive losses before auto-pause | | object | -- | Override any limit or approvalMode per platform | | number | 50 | Audit log rotation threshold (MB) | | number | 5 | Number of rotated log files to keep |

Live mode safety: When mode is "live", ALL limit fields must be finite positive numbers. Tigerpaw refuses to start with Infinity or missing limits in live mode. Paper mode allows relaxed limits for testing.

Trading Platforms

Configure the platform you want to use in the plugins section:

| | Platform | Config Key | Mode | Order Types | Auth | | ----------------------------------------------------------------------- | ------------------------------- | ------------ | --------------------- | ---------------------------------------- | ------------ | | | Alpaca (stocks) | alpaca | paper / live | market, limit, stop, stop_limit, bracket | API Key | | | Polymarket (prediction markets) | polymarket | live | limit | HMAC-SHA256 | | | Kalshi (event contracts) | kalshi | demo / live | market, limit | RSA-SHA256 | | | Manifold (play money) | manifold | play money only | market (implicit) | Bearer token | | | Coinbase (crypto) | coinbase | sandbox / live | market, limit, stop_limit | ES256 JWT | | | Interactive Brokers | ibkr | paper / live | MKT, LMT, STP, STP_LIMIT, bracket | Session | | | Binance (crypto) | binance | testnet / live | MARKET, LIMIT, STOP_LOSS_LIMIT, OCO | HMAC-SHA256 | | | Kraken (crypto + margin) | kraken | live | market, limit, stop-loss + leverage | HMAC-SHA512 | | | dYdX (perpetuals) | dydx | testnet / mainnet | market, limit (read-only) | Cosmos SDK |

Example (Alpaca):

{
  "plugins": {
    "alpaca": {
      "apiKeyId": "${ALPACA_API_KEY_ID}",
      "apiSecretKey": "${ALPACA_API_SECRET_KEY}",
      "mode": "paper"
    }
  }
}

See GETTING_STARTED.md for all platform configurations.

Risk Tiers

| Parameter | Conservative | Moderate | Aggressive | | ---------------------- | ------------ | ------------- | ---------- | | Approval Mode | Manual | Confirm (30s) | Auto | | Max Daily Spend | $100 | $500 | $2,000 | | Max Single Trade | $25 | $100 | $500 | | Max Trades/Day | 10 | 25 | 50 | | Max Open Positions | 3 | 8 | 20 | | Risk Per Trade | 1% | 2% | 5% | | Single Position Cap | 5% | 10% | 15% | | Daily Loss Limit | 3% | 5% | 10% | | Portfolio Drawdown | 10% | 20% | 30% | | Cooldown | 60s | 30s | 10s | | Consecutive Loss Pause | 3 | 5 | 8 |

Set "tier": "custom" to define your own limits without using a preset.

Per-Extension Overrides

Override limits or approval mode for individual platforms:

{
  "policy": {
    "tier": "moderate",
    "perExtension": {
      "alpaca": { "maxSingleTradeUsd": 200, "approvalMode": "manual" },
      "polymarket": { "maxDailySpendUsd": 50, "maxOpenPositions": 5 },
      "manifold": { "approvalMode": "auto" }
    }
  }
}

Any limit field or approvalMode can be overridden per platform. Unset fields inherit from the global policy.

Approval Modes

Auto -- Orders within limits execute immediately. Best for paper mode or aggressive tier.
Confirm -- Confirmation popup in the UI. Configurable timeout (default 30s for moderate tier, 15s for others) and timeout action ("deny" by default — order is rejected if not confirmed in time). Set via confirm.timeoutMs and confirm.timeoutAction.
Manual -- Every trade requires explicit operator approval. Configurable timeout (default 5 minutes) and timeout action ("deny" by default). Set via manual.timeoutMs and manual.timeoutAction.

Kill Switch

Activate via the dashboard UI (kill switch button) or from any agent channel:

"Stop all trading" → AI calls trading_killswitch_activate
"Resume trading"   → AI calls trading_killswitch_deactivate

Two modes:

Hard (default) -- Blocks ALL trading: buys, sells, and cancels all denied
Soft -- Allows sells and cancels (position exit only); blocks new buys

Auto-activates when any of these thresholds are breached:

Daily loss >= dailyLossLimitPercent
Portfolio drawdown >= maxPortfolioDrawdownPercent
Consecutive losses >= consecutiveLossPause

Pre-Trade Validation Pipeline

Every order passes through 12 sequential checks before execution. The first failure denies the order.

| # | Check | What It Validates | | --- | ---------------------- | ---------------------------------------------------------- | | 0 | Kill Switch | Is global or platform kill switch active? | | 1 | Kill Switch Auto | Should kill switch auto-activate based on current state? | | 2 | Numeric Sanity | Are order fields valid (finite, non-zero notional on buy)? | | 3 | Cooldown | Has enough time passed since the last trade? | | 4 | Balance Check | Does this trade risk more than maxRiskPerTradePercent? | | 5 | Per-Trade Size | Does notional exceed maxSingleTradeUsd? | | 6 | Daily Loss | Has daily loss reached dailyLossLimitPercent? | | 7 | Position Concentration | Would this exceed maxSinglePositionPercent in one asset? | | 8 | Max Open Positions | Are we at the maxOpenPositions cap? | | 9 | Max Trades/Day | Have we hit maxTradesPerDay? | | 10 | Daily Spend | Would cumulative spend exceed maxDailySpendUsd? | | 11 | Consecutive Losses | Have we hit consecutiveLossPause losing trades in a row? |

Audit Log

Every trade decision is logged to ~/.tigerpaw/trading/audit.jsonl with HMAC-SHA256 chain linking for tamper evidence. The log rotates at 50 MB (configurable) and keeps 5 archived files. Each entry records the action, actor, order snapshot, policy snapshot, and a chain hash linking to the previous entry.

Order Execution

The Control UI includes order entry forms on each platform page. Orders are submitted via the gateway's /tools/invoke HTTP endpoint, which:

Resolves the correct extension tool (e.g., alpaca_place_order)
Runs the order through the 12-step policy validation pipeline
Applies the configured approval mode (auto/confirm/manual)
Logs the decision to the tamper-evident audit log
Returns the result to the UI

Orders can also be placed via agent channels -- Jarvis can invoke trading tools from any connected platform, subject to the same policy gates.

How Agent Channels Work

Agent channels are how you talk to Tigerpaw from your normal chat apps. Each connected platform (Discord, Telegram, Slack, etc.) runs a lightweight agent that bridges your messages to Jarvis:

You message the agent from your normal chat app (e.g., DM the Tigerpaw agent on Discord)
Jarvis processes your request locally on your machine -- trading queries, workflow triggers, or general conversation
Responses flow back through the same channel -- alerts, trade confirmations, portfolio summaries, and more

All processing stays on your machine. The agent is just a relay -- your data never passes through third-party servers (beyond the chat platform itself). You can connect multiple channels simultaneously and they all feed into the same Message Hub.

CLI Commands

tigerpaw gateway run              # Start gateway
tigerpaw channels list            # List channels
tigerpaw channels status --probe  # Health check
tigerpaw doctor                   # Diagnostics + security audit
tigerpaw status --all             # Full system status
tigerpaw config get               # Show config

Control UI

The gateway serves a React dashboard at http://localhost:18789 with:

Dashboard -- Portfolio overview, daily P&L chart, extension status, and market prices
Message Hub -- Unified inbox across all agent channels with search, filtering, and date grouping
Assistant (Jarvis) -- AI assistant with tasks, reminders, daily briefings, and knowledge retrieval
Trading Hub -- Positions, trade history, approval queue, and risk gauges
Platform Pages -- Dedicated pages for each of the 9 trading platforms with TradingView charts (collapsible), order entry forms, and platform-specific data
Agent Channels -- Manage 20+ messaging platform agents with live status
Security -- Audit findings, credential rotation status, extension permissions
Workflows -- Visual drag-and-drop automation builder for trading events, cron schedules, and message routing
MCP -- Connect external tool servers and expose Tigerpaw tools to other AI agents
Models -- Local LLM management, provider configuration, model health monitoring
Settings -- Risk tier selection, approval mode, per-extension overrides, remote access configuration
Config -- JSON config editor with live validation

Remote Dashboard Access

By default the dashboard is only accessible on the machine running Tigerpaw. To access it from other devices:

| Method | Encryption | Install Required | Best For | | ------------------------ | --------------------------------- | ---------------------------- | ------------------------- | | Local only (default) | N/A | None | Single-machine use | | Tailscale | End-to-end (WireGuard) | Server + every client device | Privacy-sensitive setups | | Cloudflare Tunnel | TLS (Cloudflare decrypts at edge) | Server only (cloudflared) | Easy access from anywhere |

Configure in Settings > Dashboard Access or during first-run setup (tigerpaw start). See GETTING_STARTED.md for step-by-step instructions.

Your API keys never leave your machine regardless of access mode. Remote access only exposes the dashboard UI -- trade execution and credential storage remain local.

Trading Agent Commands

The trading-commands extension provides 8 tools accessible from any connected agent channel:

| Command | Description | | ------------------------------- | ------------------------------------------------ | | trading_portfolio_summary | Portfolio value per platform, total, drawdown | | trading_daily_metrics | P&L, spend, trades, consecutive losses vs limits | | trading_positions | All open positions with value and % allocation | | trading_killswitch_status | Global + per-platform kill switch state | | trading_killswitch_activate | Halt all trading with reason | | trading_killswitch_deactivate | Resume trading | | trading_risk_status | % utilization of each limit with progress bars | | trading_recent_trades | Recent trade decisions from the audit log |

These work from Telegram, Discord, Slack, or any connected agent channel -- ask Jarvis "What's my portfolio?" or "Stop all trading" and it invokes the right tool.

Notifications

Tigerpaw sends real-time notifications for trading events -- entirely local, no external services:

In-app toasts -- Color-coded alerts for order approvals, denials, kill switch changes, and limit warnings
Notification bell -- Badge counter in the dashboard header showing undismissed notifications
Browser notifications (opt-in) -- Desktop notifications via the browser Notification API when the dashboard tab is open
Per-platform filtering -- Enable or disable notifications per trading platform (e.g. only Polymarket alerts)

Events tracked:

| Event | Severity | Description | | ----------------------- | -------- | -------------------------------------------- | | Order approved | Success | Order passed all policy checks | | Order denied | Error | Order blocked by policy engine (with reason) | | Order pending | Info | Awaiting manual/confirm approval | | Kill switch activated | Error | Trading halted (with trigger reason) | | Kill switch deactivated | Success | Trading resumed | | Limit warning | Warning | Approaching 80% of a configured limit |

Per-Platform Filtering

In Settings > Notifications, toggle notifications on or off for each trading platform independently. For example, enable notifications only for Polymarket and disable all others. Global events (kill switch activation, limit warnings without a specific platform) are always shown regardless of filter settings.

Proactive Agent Channel Notifications

Optionally push trading alerts to any agent channel (Telegram, Discord, Slack, etc.):

{
  "trading": {
    "notifications": {
      "enabled": true,
      "targets": [
        { "channel": "telegram", "to": "123456789" },
        {
          "channel": "discord",
          "to": "channel-id",
          "events": ["trading.order.denied", "trading.killswitch.activated"]
        }
      ]
    }
  }
}

Each target can filter which events it receives via the events array. Omit events to receive all trading notifications. Supported events: trading.order.approved, trading.order.denied, trading.order.pending, trading.order.submitted, trading.order.filled, trading.order.failed, trading.killswitch.activated, trading.killswitch.deactivated, trading.limit.warning.

No remote alerts are sent by default. Notifications stay on your machine unless you configure the above.

Development

pnpm build          # Compile TypeScript
pnpm ui:build       # Build React UI
pnpm check          # Lint + format
pnpm test:fast      # Unit tests
pnpm test           # All tests

Troubleshooting

| Problem | Fix | | --------------------------------- | -------------------------------------------------------------------------------------------------------------------------- | | Missing config error on startup | Run tigerpaw start (auto-creates config) or tigerpaw setup | | Dashboard shows blank page | Clear browser cache, check http://localhost:18789 is accessible | | Trades not executing | Check tigerpaw doctor — verify the platform is connected and trading is enabled | | Kill switch stuck on | Run tigerpaw trading resume or click the kill switch button in the dashboard | | Port already in use | Another instance may be running — check with lsof -i :18789 or change port via tigerpaw config set gateway.port <port> | | openclaw commands not found | Use tigerpaw instead — the legacy openclaw alias works only if the old package was installed |

File Locations

| Purpose | Path | | --------------- | --------------------------- | | Config | ~/.tigerpaw/tigerpaw.json | | SQLite database | ~/.tigerpaw/tigerpaw.db | | Sessions | ~/.tigerpaw/sessions/ |

The SQLite database (tigerpaw.db) stores workflows, credentials (encrypted at rest), execution history, the trading audit log, and policy state. To back up your data, copy this single file while the server is stopped.

Upgrading from flat files? On first startup after upgrade, Tigerpaw automatically imports existing data from credentials/, workflows/, workflow-runs/, workflow-versions/, and trading/ into the SQLite database. The original directories are renamed with a .pre-sqlite-backup suffix as a safety measure.

Security

Policy-gated trading: Every order passes through 12 pre-trade checks; extensions fail-safe (block orders) when the policy engine is unavailable
HMAC-signed requests: API secrets are never sent as plaintext headers
Tamper-evident audit log: HMAC-SHA256 chain-linked JSONL for every trade decision
Kill switch: Instant halt with auto-activation on limit breach

See SECURITY.md for the full security architecture and vulnerability disclosure policy.

Contributing

See CONTRIBUTING.md.

Acknowledgments

Built on OpenClaw by Peter Steinberger. Tigerpaw includes all of OpenClaw's 40+ messaging channel integrations and agent runtime, and extends them with a trading engine, gateway security hardening, a React 19 dashboard, real-time notifications, and Docker multi-arch images. Grateful to Peter and the OpenClaw community for the foundation.

License

Dual-licensed. Code originating from OpenClaw remains under MIT. All Tigerpaw additions are licensed under AGPL-3.0. See LICENSE for details.