Your always-on personal AI assistant. One pure-Swift daemon on hardware you own.

clawd pairs a private Telegram bot with the LLM of your choice. It remembers what you tell it and runs scheduled and proactive tasks. Consequential tool calls wait for your approval. Everything it keeps stays in one directory on your own machine: a SQLite database, encrypted secret envelopes, and Markdown files you edit by hand.

Features

• One binary. Your machine. Run your always-on personal assistant as a single Swift daemon, with persistent state stored on hardware you own.
• Coding tasks from chat. Opt in to Coder to delegate an approved task from your DM or a configured group topic to your native Codex installation, then receive a structured result card when the background job finishes. Local changes and GitHub pull requests use your installed tools and repository rights.
• Scheduled tasks that learn from feedback. Opt in with CLAW_LEARNING_ENABLED=true. Correct one result to start a bounded lesson trial; two positive runs can promote it. Inspect lessons and roll back a promotion with /learning.
• Proactive, on your clock. "Every weekday at 07:00" schedules fire once per occurrence across restarts and DST changes, and an opt-in heartbeat respects quiet hours.
• Daily notes across sessions. Set CLAW_JOURNAL_ENABLED=true for your personal deployment. clawd batches decisions, results and unfinished work into dated Markdown files and loads today's and yesterday's notes after /new or a restart. Inspect or delete them with /journal. Daily journals explains delays and extra model usage.
• Durable memory. Facts you confirm persist in SQLite, and clawd recalls them by importance and recency. Workspace Markdown files hold your profile and curated memory, and conversation history is full-text searchable.
• Tools behind a policy engine. web_fetch sits behind an SSRF gate; writes and code execution wait for an explicit tap-to-approve in Telegram. clawd enforces policy in code and treats inbound content as data, never as instructions.
• Bring your own model. Any OpenAI-compatible endpoint works, and clawd auth login can run an eligible model on a ChatGPT subscription.
• Skills you write once. A skills/<name>/SKILL.md file shows up in context as its name and one line about when to use it; when a task matches, clawd loads the body and follows your procedure instead of asking you to paste it again. Send /skills to see every accepted skill and each file the scanner rejected.
• Tools from MCP servers. List a server, store its token encrypted, and its tools join the built-ins as the least-trusted tools clawd has. Calls ask by default; you may mark a named tool safe, but the exfiltration gate can still require approval. Only you can add a server or change what it exposes.
• A real Telegram chat. Answers stream in as live message drafts. Tap Stop to cancel that turn and keep queued messages. /stop cancels current and queued turns; /new starts a fresh session. clawd transcribes voice notes on-device (macOS 26), and it looks at photos you send if your model can see them.
• Sandboxed code execution. Untrusted code runs in a fresh disposable VM per request (macOS 26 arm64, off by default).

The approval card

A file write suspends the run until you answer. Every field on the card comes from the daemon's own record of the action: the target path after symlink and .. resolution, the size, and a preview of the content. Coder uses a dedicated card with the complete source, workspace, publication scope, provided task or selected issue, and optional additional requirements; its completion card puts the outcome first and keeps technical evidence compact. Tap Deny and clawd writes nothing.

Install

curl -fsSL https://raw.githubusercontent.com/ivan-magda/swift-claw/main/install.sh | sh

Everything lands in ~/.swift-claw, with no sudo. The script verifies every download against the release checksums, stages the service files, and prints the next steps. Pin a release with curl … | CLAWD_VERSION=v0.2.0 sh, read the script source first, or follow the manual route in docs/INSTALL.md (macOS 26+ arm64; Linux x86_64 with glibc 2.38+).

Or build from source with the pinned Swift 6.4 toolchain. Linux source builds also need libsqlite3-dev.

Quick start

1. Get a bot token from @BotFather (send /newbot).
2. Edit ~/.swift-claw/clawd.env: set the token and your LLM provider (CLAW_LLM_BASE_URL, CLAW_LLM_MODEL, CLAW_LLM_API_KEY).
3. Load the config and encrypt your secrets at rest: set -a && . ~/.swift-claw/clawd.env && set +a && clawd secrets seal
4. Say hello once in the foreground: clawd run, then send /start to your bot. The refusal shows your numeric ID; set it as CLAW_ALLOWLIST=<id> in clawd.env, then Ctrl-C.
5. Check health and start the service: set -a && . ~/.swift-claw/clawd.env && set +a && clawd doctor, then run the start command doctor prints.

The full walkthrough, including the ChatGPT-subscription route and troubleshooting, is in docs/GETTING_STARTED.md.

In Telegram, /schedule creates or lists jobs; /runnow <jobId> runs one now; /pause, /resume, and /cancel control a job. /learning <jobId> shows its lessons and trial, and /learning reset <jobId> asks you to confirm an empty lesson set. /help lists commands and confirmation rules. /journal, /journal show YYYY-MM-DD, and /journal delete YYYY-MM-DD inspect and manage personal daily notes, including while automatic journaling is disabled.

Security model

swift-claw assumes you are the only person it serves in its normal personal deployment. Configured groups are a supervised exception: use a separate nonpersonal state root, trust the participants, and understand that their ordinary tool approvals are relaxed; see group Coder configuration.

• Default-deny. Only allowlisted Telegram IDs get a conversation. clawd refuses everyone else, and answers /start with the sender's own numeric ID so you can allowlist them. CLAW_ALLOWLIST only ever adds, so revoking an ID means deleting its row from the database (details).
• Secrets encrypted at rest. clawd secrets seal wraps the bot token and API keys in an AES-GCM envelope. Plaintext env secrets remain available as a dev fallback that warns on every boot.
• Approvals are durable and bound to their prompt. In your DM, file writes, memory writes, code execution, and native Coder submissions suspend into a durable state machine until you decide. In a configured group, Coder submission is the one action that always asks: any current participant can approve or deny from its original approval message, and clawd checks membership with Telegram at the tap. Pending approvals expire to deny.
• Prompt injection contained in the personal deployment. Messages, web content, tool output, and stored memory enter the context as untrusted data. Once a session has both ingested untrusted content and pulled your private files into context, fetching an arbitrary URL also needs your approval. clawd pins your LLM and search providers in config, and the model cannot redirect them.

Opt-in daily journals are private labeled data that preserve useful conclusions without setting session taint. While enabled, clawd checks outbound tool arguments for MEMORY/USER substrings independently of taint. This check adds no approvals and does not prevent every possible disclosure.

The full model is in docs/ARCHITECTURE.md (§12). To report a vulnerability, see SECURITY.md.

Customize your agent

Persona and behavior live in Markdown files under ~/.swift-claw/workspace/:

File	Shapes	Trust
SOUL.md	Personality and tone	System prompt
AGENTS.md	Behavior rules	System prompt
TOOLS.md	When and how to use tools	System prompt
USER.md	Who you are	Untrusted, labeled
HEARTBEAT.md	The proactive heartbeat checklist	Heartbeat runs only
skills/<name>/SKILL.md	A procedure the agent loads when a task calls for it	Untrusted, labeled

MCP servers go in ~/.swift-claw/mcp.yaml, with their tokens stored encrypted by clawd mcp set-token. Other runtime knobs are environment variables: the model route (CLAW_LLM_MODEL), an optional fallback route (CLAW_LLM_FALLBACK_MODEL, off unless you set it), USD budgets, schedules and quiet hours, voice locales, sandbox limits. .env.example documents every variable; docs/CUSTOMIZATION.md is the guide.

Documentation

You want to	Read
Set it up end to end	docs/GETTING_STARTED.md
Install, update, or uninstall	docs/INSTALL.md
Make it yours	docs/CUSTOMIZATION.md
Run it as a service	docs/INSTALL.md
Develop and test locally	docs/LOCAL_DEV.md
Follow the Swift style baseline	docs/CODE_STYLE.md
Understand the design	docs/ARCHITECTURE.md
See the architecture as diagrams	swift-claw-architecture.html (source)
Report a vulnerability	SECURITY.md

Experimental benchmarks, corpora and the Swift evaluation harness live in swift-claw-evals. The daemon's scheduled-learning implementation and product tests stay in this repository. Building and testing clawd does not require the laboratory.

Contributing

Contributions are welcome. Open an issue to discuss what you have in mind before sending a pull request; CONTRIBUTING.md has the details and the lint/test gate.

Please follow our Code of Conduct when participating in the project.

License

MIT © Ivan Magda