⚒ Clawmark™ Registry · 34 Tools

Every RelayForge tool —
scored on the Clawmark standard.

Six criteria — Classification, Developer Accountability, Test Documentation, Independent Validation, and Lobster Usability (split into Discovery and Reliability) — each scored 0–5. The aggregate is their mean, on a 0–5 claw scale. This page renders live from the ARIA registry; nothing here is hardcoded.

What does this mean? Read the rubric →
34 tools

CalClaw

operatorneeds scope

4.0 / 5

Beta

Deterministic NAMUR NE43 calibration data auditor for industrial instruments. Accepts as-found/as-left data for up to 500 records per request, applies NE43 pass/fail rules with full error calculations, flags pull-from-service violations, and persists results to Supabase for historical trending. No AI backbone — pure math against a published standard. Includes an operator web dashboard.

CertClaw

operatorneeds scope

3.8 / 5

Certified

Personal certification and CEU tracker. Stores certificates, expiry windows, and continuing-education credits with reminders. Beta upstream-auth-only deployment.

IndexClaw

operatorneeds scope

3.7 / 5

Beta

Entity extraction engine for industrial engineering drawings. Auto-detects document type (P&ID, SLD, cable schedule, loop diagram, cause and effect, isometric, equipment GA) and extracts the correct entity set for each type: ISA instrument tags, equipment tags, line numbers, breaker tags, bus IDs, protection relay tags, cable numbers, and terminal block points. Includes a CMMS diff engine to identify gaps between drawing data and Maximo/SAP PM/generic CSV exports. Flags pages requiring a vision model pass. Persists results to Supabase with JSONB storage and GIN index for fast tag search.

StreamClaw

agentneeds scope

3.7 / 5

Certified

Personalized free streaming guide for Pluto TV, Tubi, Amazon Freevee, Peacock Free, and other FAST services. Supports up to 8 family profiles with individual taste models, content rating filters, and watch history. Features live channel grid with real-time progress bars, obscurity scoring for hidden-gem discovery, watchlist, Gemini AI content summaries, ElevenLabs TTS for hands-free browsing, and device-aware deep links. Exposes MCP tools via /mcp router: get_live_guide, get_recommendations, record_preference, search_content. SSE streaming endpoint available for real-time recommendation cards.

Time

freesafe

3.5 / 5

CertifiedGrandfathered

Returns current UTC time and allows timezone conversion. No network calls — pure system clock.

ParseClaw

operatorneeds scope

3.5 / 5

Beta

PDF ingestion pipeline for industrial instrumentation manuals. Pulls PDFs from Cloudflare R2, extracts numbered procedure steps, WARNING/CAUTION/NOTE safety blocks, torque/pressure specs, ISA instrument tags, part numbers, and cross-document references. OCR fallback via pytesseract for scanned pages. Returns confidence scoring and epistemic metadata on every result. Persists structured chunks to Supabase for downstream RAG search.

SpecClaw

operatorneeds scope

3.5 / 5

Beta

Multi-domain specification validation engine for industrial instrumentation. Validates instrument datasheets (PDF via R2 or raw text) against caller-supplied requirements using deterministic rules. Runs automatic domain checks across four domains: I&E (area classification, SIL rating), commercial electrical (NEC compliance), mechanical piping (ASME ratings), and general contractor. Returns pass/fail/warning/insufficient per requirement with severity and epistemic summaries. Includes a QuickCheck endpoint for pre-procurement checklists. Full audit trail in Supabase.

FieldClaw

freesafe

3.5 / 5

Beta

RelayForge FieldClaw — field-operations product (OS tier). Hybrid Express + FastAPI application. Registered pending Clawmark review.

SafetyClaw

operatorneeds scope

3.5 / 5

Beta

Incident-aware JHA/JSA workbench for industrial maintenance crews. Generates step-by-step hazard analyses against OSHA 3071 / API RP 75 / ANSI Z10 or custom templates, annotates each step with matched real-world incidents (CSB/OSHA/enterprise), and produces signed receipts. Warns. Never recommends — every output requires competent-person review; recurring jobs force re-review every 90 days.

ParkClaw

freesafe

3.3 / 5

Beta

Park conditions and trip-planning tool for U.S. national and state parks. Aggregates NPS directory data, NWS weather, InciWeb fire incidents, and AirNow air quality. Serves an MCP discovery manifest at /.well-known/mcp-tool.

CampClaw

freeneeds scope

3.2 / 5

Beta

Campground discovery and reservation watcher. Pulls Recreation.gov, ReserveCalifornia, and state-park availability; supports cancellation watchers with optional auto-claim.

Memory

freesafe

3.0 / 5

BetaGrandfathered

Persistent key-value knowledge graph. Agents store and recall facts across sessions via a local graph store. Core to lobster identity.

LearnClaw

agentneeds scope

3.0 / 5

Beta

Adaptive learning companion PWA with per-user cognitive profiling. Onboards learners via a dynamic question bank across logistics, frustration-style, and troubleshooting domains to build a persistent Learner Profile. Provides a Socratic Terminal tutor, AI-generated flashcard decks with SM-2 spaced repetition, adaptive quiz engine with difficulty scaling, action planner across daily/weekly/monthly/quarterly horizons, and StoryTime for younger learners. Backed by Gemini 2.5 Flash. Privacy-first: all learner data stored client-side. MCP integration in development.

Sequential Thinking

agentsafe

2.8 / 5

BetaGrandfathered

Structured multi-step reasoning tool. Breaks problems into sequential thought chains before answering. Improves accuracy on complex tasks.

MealClaw

freeneeds scope

2.8 / 5

Beta

AI-powered pantry and recipe management. Tracks household inventory, suggests recipes from on-hand ingredients, manages expiry windows. Larry voice assistant for in-kitchen capture.

WildClaw

freesafe

2.8 / 5

Beta

Wildlife observation and identification tool. Aggregates iNaturalist, eBird, and NPS species data for location-aware species lists, recent sightings, and seasonal patterns.

NarrateClaw

freeneeds scope

2.8 / 5

Beta

Long-form audio narration with chapter/scene markers. Turns text or markdown into a paced audio experience with optional voice selection.

SerialClaw

freeneeds scope

2.8 / 5

Beta

Reading-coach companion for serialized fiction and long-form text. Tracks comprehension, surfaces context recaps, and provides per-chapter check-ins via a friendly Coach persona.

LadderLab

freesafe

2.8 / 5

Beta

Programmable Logic Controller (PLC) ladder-logic tutor. Combines an AI tutor with the Rung Climber drill engine for visual-style ladder-logic instruction. Production deployed.

CrewClaw

operatorneeds scope

2.8 / 5

Alpha

Private multi-tenant crew operating system: shared calendar with two-way Google sync, document vault, memoir recording, safety alerts, medication tracking, shared lists, and a trade Pro Toolbox (calibration analysis, nameplate OCR, shift-sync, lineman logbook, FTO narratives). Own subscription OS product.

Git (Read-Only)

assistantsafe

2.7 / 5

BetaGrandfathered

Read-only git operations: log, diff, status, show. Cannot commit, push, or modify the working tree.

SQLite (Read-Only)

assistantsafe

2.7 / 5

BetaGrandfathered

Read-only access to SQLite databases: SELECT queries only. Cannot INSERT, UPDATE, DELETE, or modify schema.

Filesystem (Scoped)

assistantneeds scope

2.5 / 5

BetaGrandfathered

Read/write files within a strict agent-scoped directory allowlist. Cannot access paths outside the agent's assigned workspace.

Fetch (Scoped)

assistantneeds scope

2.5 / 5

BetaGrandfathered

HTTP GET requests to allowed domains only. Scoped allowlist prevents SSRF and access to internal Railway hostnames.

GitHub

operatorneeds scope

2.5 / 5

BetaGrandfathered

Full GitHub API access: repos, issues, PRs, commits, code search. Requires GitHub personal access token. Read+write.

Everything (Test/Demo)

agentneeds scope

2.2 / 5

BetaGrandfathered

Reference MCP server exposing all MCP primitives: prompts, resources, and tools. Used for testing and protocol validation.

PostgreSQL

operatorunsafe

2.2 / 5

BetaGrandfathered

Full PostgreSQL access via connection string. Read and write queries. Operator tier — direct DB access. Connection string required.

Puppeteer

agentneeds scope

2.0 / 5

BetaGrandfathered

Browser automation via Puppeteer: screenshot, navigate, click, fill forms. Runs headless Chrome. Requires system Chrome install.

DinoClaw

freesafe

2.0 / 5

Beta

Dinosaur fact and discovery explorer. Aggregates paleontology data for taxon lookups, fossil sites, and child-friendly fact bundles.

Google Drive

agentneeds scope

1.8 / 5

AlphaGrandfathered

Read and search Google Drive files via OAuth. Lists files, reads content, searches by query. Requires Google OAuth authorization.

Gmail (Read-Only)

agentneeds scope

1.8 / 5

AlphaGrandfathered

Read Gmail messages, search inbox, list threads. Cannot send or modify messages. Requires Google OAuth authorization.

Google Calendar (Read-Only)

agentneeds scope

1.8 / 5

AlphaGrandfathered

Read Google Calendar events, list upcoming appointments, check free/busy. Cannot create or modify events. Requires Google OAuth authorization.

Google Workspace

agentneeds scope

1.8 / 5

AlphaGrandfathered

Full Google Workspace integration: Drive, Gmail, Calendar, Docs, Sheets. Unified OAuth credential covers all Google services. Agent tier — send capability included.

The Clawmark Standard — 5 criteria, 6 sub-scores

Each criterion is scored 0–5. The aggregate is the mean of all six sub-scores, displayed as a 0–5 claw rating. Status is derived from the aggregate. Read the full rubric →

C1Classification

Category, input/output types, permission scopes, and a machine-readable manifest at /api/v1/spec.

C2Developer Accountability

A MAINTAINER.md with a named author, version history, issue contact, and support SLA.

C3Test Documentation

A real test suite covering unit and API routes, with edge cases and documented coverage gaps.

C4Independent Validation

Review by someone other than the author — DP audit, agent QA, or external security review + red-team.

C5aLobster Usability — Discovery

A lobster can discover and call the tool via ARIA + the manifest, with standard errors, no human help.

C5bLobster Usability — Reliability

The tool behaves under real agent use: failure recovery, multi-call chains, timeouts, epistemic metadata.

Full rubric →Trust standards →Architecture docs →PrivacyTermsCookies