{"id":"qwibitai-nanoclaw","name":"nanoclaw","homepage":"https://nanoclaw.dev","repo_url":"https://github.com/qwibitai/nanoclaw","category":"ai-ml","subcategories":[],"tags":["ai-agents","claude","container-sandboxing","messaging","self-hosted","scheduled-jobs","typescript","openclaw-alternative"],"what_it_does":"NanoClaw is a self-hosted personal AI assistant that orchestrates Claude Agent SDK-based agents inside isolated Linux containers, providing multi-channel messaging (e.g., WhatsApp/Telegram/Slack/Discord/Gmail), per-group memory and filesystem isolation, and scheduled/recurring tasks. It’s designed to be customized by modifying code and by adding “skills” (Claude Code skills) rather than using configuration-heavy setups.","use_cases":["Run a private Claude-based assistant with isolated per-group context and memory","Automate recurring tasks (e.g., weekly briefings, scheduled updates) that can message you","Connect an assistant to one or more messaging channels (WhatsApp/Telegram/Slack/Discord/Gmail) via skills","Give agents controlled access to mounted directories (e.g., an Obsidian vault folder) for retrieval and writing","Use Claude Agent SDK containerized execution for better host isolation than in-process permission checks"],"not_for":["A managed SaaS you can drop into without hosting or container/runtime setup","A turnkey enterprise contact-center or fully multi-tenant platform with centralized admin/monitoring","Use cases requiring a stable, documented external HTTP API surface for programmatic integration (it appears to be primarily an agent runtime/CLI workflow)"],"best_when":"You want to self-host an agent runtime with container filesystem isolation and you’re comfortable using Claude Code skills and customizing a fork.","avoid_when":"You need a stable public API/SDK for third-party automation, or you cannot operate Docker/containers and handle local secrets/configuration via environment variables and related tooling.","alternatives":["OpenClaw","Self-hosted Claude/LLM agent frameworks with container execution (e.g., generic agent runners)","RPA/workflow tools for scheduled messaging instead of agent-based orchestration"],"af_score":28.2,"security_score":61.2,"reliability_score":36.2,"package_type":"skill","discovery_source":["openclaw"],"priority":"high","status":"evaluated","version_evaluated":null,"last_evaluated":"2026-03-29T13:08:23.419095+00:00","interface":{"has_rest_api":false,"has_graphql":false,"has_grpc":false,"has_mcp_server":false,"mcp_server_url":null,"has_sdk":true,"sdk_languages":["TypeScript","JavaScript"],"openapi_spec_url":null,"webhooks":false},"auth":{"methods":["Environment variables for Anthropic-compatible model endpoints (ANTHROPIC_BASE_URL, ANTHROPIC_AUTH_TOKEN)","Channel credentials managed via skills and a vault/proxy for outbound requests (OneCLI Agent Vault mentioned)"],"oauth":false,"scopes":false,"notes":"Authentication appears to be primarily local/secrets-based for model endpoints and per-channel credentials via the documented skills/setup flow; outbound requests reportedly route through OneCLI’s Agent Vault for per-agent policies and rate limits."},"pricing":{"model":null,"free_tier_exists":false,"free_tier_limits":null,"paid_tiers":[],"requires_credit_card":false,"estimated_workload_costs":null,"notes":"README describes self-hosting; no pricing tiers are provided in the supplied materials. Cost will depend on your LLM usage and messaging provider costs."},"requirements":{"requires_signup":false,"requires_credit_card":false,"domain_verification":false,"data_residency":[],"compliance":[],"min_contract":null},"agent_readiness":{"af_score":28.2,"security_score":61.2,"reliability_score":36.2,"mcp_server_quality":0.0,"documentation_accuracy":55.0,"error_message_quality":0.0,"error_message_notes":null,"auth_complexity":55.0,"rate_limit_clarity":45.0,"tls_enforcement":70.0,"auth_strength":70.0,"scope_granularity":45.0,"dependency_hygiene":45.0,"secret_handling":70.0,"security_notes":"README claims agents run in Linux containers with filesystem isolation and explicit mounts, and that outbound credentials do not enter containers (routed via OneCLI’s Agent Vault with per-agent policies and rate limits). However, TLS enforcement for all network paths and concrete authentication/authorization scope granularity are not evidenced in the provided materials. Dependency security posture cannot be verified from the given manifest alone (only a few dependencies are listed).","uptime_documented":0.0,"version_stability":55.0,"breaking_changes_history":50.0,"error_recovery":40.0,"idempotency_support":"false","idempotency_notes":null,"pagination_style":"none","retry_guidance_documented":false,"known_agent_gotchas":["This project is primarily operated via Claude Code skills/CLI flow; integration as a generic automated API may require additional work.","Skills modify a fork; agent workflows may be sensitive to local skill configuration and mounted directory permissions."]}}