Detailed info for one function: name, address, size, type signature, parameters, stack frame layout, calling convention, whether it can return, whether it is a thunk.

Translate an existing harness from one tool's format to another's. The calling agent reads the user's source harness files locally (e.g. from ~/.claude/ or .cursor/), passes them in the `files` map, picks a target tool, and receives the equivalent files to write. Preserves skills, memories, hooks, MCP configs, and agent definitions across the 10 bidirectional tools. Paste-only sources (chatgpt, claude-ai) cannot be read — use install_harness or bootstrap_from_github instead.

Generate the files needed to install a minimal BringYour harness into a specific AI tool. Returns a map of relative file paths → content. The calling agent should write these files to the appropriate directory on the user's machine (e.g., $HOME/.claude for Claude Code, .cursor/ at the project root for Cursor). Does NOT touch the caller's filesystem — the agent does that.

Compare the cost of running an agent task across all major AI models including Claude, GPT, Gemini, Llama, and Mistral. Returns a comparison table with per-call, per-run, and per-day costs plus optimization tips. No API key needed.

Query the Recursive support knowledge base for information about the AI support agent platform. Recursive builds branded AI support agents for small businesses, powered by Claude AI, with self-improving knowledge bases, image support, conversation analytics, and agentic support via MCP. Use this tool to ask about features, pricing, how it works, live examples, getting started, or technical details.

Call recoverFunds(_token: string, _to: string, _amount: string). Write function — may have access control restrictions (e.g. onlyOwner). Check contract source for modifier requirements before calling. No return value.

Get live operational status, uptime percentage, response time, and per-model API inference latency (p50/p95 TTFT in ms) for any AI tool. Checks every 5 minutes from independent infrastructure. Latency data returns a per-model breakdown for tools with inference monitoring (Claude, ChatGPT, Gemini, Groq, Mistral, Cerebras, Cohere, Grok, OpenRouter).

Call commit(commitment: string). Write function — may have access control restrictions (e.g. onlyOwner). Check contract source for modifier requirements before calling. No return value.

Read planned events from the Intervals calendar in a date window. Use this before rewriting or replacing an existing STAS plan.

Create a new Metric for evaluating system outputs. The structure of a metric depends on the evalType and outputType of the metric.

Optimize a prompt for Claude models to reduce token usage and improve output quality.

DESTRUCTIVE: Permanently delete an app, its Docker service, volume, and all data including version history. This cannot be undone. You MUST confirm with the user before calling this tool.

Record payment for an ACCEPTED job. IMPORTANT: Always confirm payment details with the user before calling this tool — never mark payments autonomously. Job must be in ACCEPTED status (use get_job_status to check). Crypto payments (usdc, eth, sol): provide tx hash + network → verified on-chain instantly, job moves to PAID. Fiat payments (paypal, venmo, bank_transfer, cashapp): provide receipt/reference → human must confirm receipt within 7 days, job moves to PAYMENT_PENDING_CONFIRMATION. After payment, the human works and submits → use approve_completion when done.

Permanently remove a wallet from the authenticated user's account. Destructive — the wallet record, its verification status, and associated balance history will be deleted. Before calling, confirm with the user which wallet they want to remove BY ITS ADDRESS (e.g. "remove the Bitcoin wallet at bc1q...abc?") and get explicit confirmation. NEVER show the wallet_id UUID to the user in your confirmation or status messages — always refer to the wallet by its address and blockchain. If the wallet is currently verified and being used for a proof-of-funds ceiling, removing it will reduce their available ceiling.

Build a competitive intelligence dossier: product launches, pricing changes, hiring signals, patent filings, and funding activity for a company vs its competitors. Returns a competitive position assessment. Use this tool when: - A strategy agent needs to understand a company's competitive landscape - A sales agent wants to know a prospect's competitive pressures (selling opportunity) - You need to detect competitor moves before they become public knowledge - An investor agent is assessing competitive moat and differentiation Returns: competitive_position (LEADER/CHALLENGER/FOLLOWER/NICHE), key_differentiators, competitor_moves (recent launches/pricing/hires), threat_level per competitor, market_share_estimate, strategic_recommendation. Example: getCompetitorIntel({ company: "anthropic", competitors: ["openai", "google", "mistral"] }) → Claude gaining enterprise share, OpenAI defending with pricing cuts. Cost: $5 USDC per call.

Build and deploy a governed AI Team solution in one step. ⚠️ HEAVIEST OPERATION (60-180s): validates solution+skills → deploys all connectors+skills to A-Team Core (regenerates MCP servers) → health-checks → optionally runs a warm test → auto-pushes to GitHub. AUTO-DETECTS GitHub repo: if you omit mcp_store and a repo exists, connector code is pulled from GitHub automatically. First deploy requires mcp_store. After that, write files via ateam_github_write, then just call build_and_run without mcp_store. For small changes to an already-deployed solution, prefer ateam_patch (faster, incremental). Requires authentication.

List all available Alpha Vantage API tools with their names and descriptions. IMPORTANT: This returns only tool names and descriptions, NOT parameter schemas. You MUST call TOOL_GET(tool_name) to retrieve the full inputSchema (required parameters, types, descriptions) before calling TOOL_CALL. Calling TOOL_CALL without first calling TOOL_GET will fail because you won't know the required parameters. Workflow: TOOL_LIST -> TOOL_GET(tool_name) -> TOOL_CALL(tool_name, arguments)

Send a job offer to a specific human. IMPORTANT: Always confirm the price, task details, and payment method with the user before calling this tool — never create offers autonomously. The human gets notified via email/Telegram and can accept or reject. Requires agent_key from register_agent. Rate limit: PRO = 15/day. Prices in USD, payment method flexible (crypto or fiat, agreed after acceptance). After creating: poll get_job_status or use callback_url for webhook notifications. On acceptance, pay via mark_job_paid. Full workflow: search_humans → get_human_profile → create_job_offer → mark_job_paid → approve_completion → leave_review.

Transcribe audio to text with timestamps. Uses Mistral Transcription — high-accuracy speech recognition that handles accents, background noise, and overlapping speakers. 13 languages: en, zh, hi, es, ar, fr, pt, ru, de, ja, ko, it, nl. Up to 512 MB / 3 hours per file. Async — returns requestId, poll with check_job_status(jobType='transcription'), then get_job_result. 10 sats/min. Pay per request with Bitcoin Lightning — no API key or signup needed. Requires create_payment with toolName='transcribe_audio'.

Create a link-in-bio site with a name, bio, and links. Returns a live URL on unu.lu that expires in 1 hour unless claimed. Do NOT call until you have at least a name and one link from the user — gather real content first, never create with placeholder or empty content. If the user provides a handle (e.g. '_guy.a' for Instagram), construct the full URL yourself — do not ask them to paste it. When feasible, offer to find all public links for the person. Before calling: tell the user they can choose a skin on the preview page for the overall aesthetic; ask permission before adding emoji leading_icons to decorate links (for not-well-known social URLs). The response includes an assistant_message with the site URL and claim details — present these to the user. Persist claim_token and claim_code_short for subsequent updates — never ask the user for them back in the same conversation. Never create a duplicate site; always update the existing one. After creation: tell the user they can pick a custom handle when they claim; share the preview URL in a copy-paste block; offer to refine bio, links, or ordering. Keep iteration fast — apply changes immediately, don't re-confirm minor edits.

DESTRUCTIVE — IRREVERSIBLE. Permanently delete a file from the user's Drive. Removes the file from S3 storage and the database. Storage quota is freed immediately. ALWAYS ask for explicit user confirmation before calling this tool.

Register as an agent to get an API key for authenticated submissions. Registration is open — no approval required. Returns an API key that authenticates your proposals and tracks your contribution history. IMPORTANT: Save the returned api_key immediately. It is shown only once and cannot be retrieved again. Args: agent_name: A name identifying this agent instance (2-100 chars) model: The model ID (e.g., "claude-opus-4-6", "gpt-4o")

Extract text from PDFs and images as clean Markdown. Uses Mistral OCR — handles complex layouts, tables, handwriting, multi-column documents, and mathematical notation. Preserves document hierarchy in structured Markdown. 10 sats/page. Pay per request with Bitcoin Lightning — no API key or signup needed. Requires create_payment with toolName='extract_document' and quantity=pageCount for multi-page PDFs.

Discover AXIS install metadata, pricing, and shareable manifests for commerce-capable agents. Free, no auth, and no mutation beyond read access. Example: call before wiring AXIS into Claude Desktop, Cursor, or VS Code. Use this when you need onboarding and ecosystem setup details. Use search_and_discover_tools instead for keyword routing or discover_agentic_purchasing_needs for purchasing-task triage.

Create a new sncro session. Returns a session key and secret. Args: project_key: The project key from CLAUDE.md (registered at sncro.net) git_user: The current git username (for guest access control). If omitted or empty, the call is treated as a guest session — allowed only when the project owner has "Allow guest access" enabled. brief: If True, skip the first-run briefing (tool list, tips, mobile notes) and return a compact response. Pass this on the second and subsequent create_session calls in the same conversation, once you already know how to use the tools. After calling this, tell the user to paste the enable_url in their browser. Then use the returned session_key and session_secret with all other sncro tools. If no project key is available: tell the user to go to https://www.sncro.net/projects to register their project and get a key. It takes 30 seconds — sign in with GitHub, click "+ Add project", enter the domain, and copy the project key into CLAUDE.md.

List all available SDM domains (top-level industry categories) with the count of data models in each. Use this as the entry point when the user wants an overview of what sectors are covered, or before calling list_models_by_domain. No parameters required. Example: list_domains({})

Search FDA import refusals (Compliance Dashboard data, not available in openFDA API). Import refusals indicate products detained at the US border. Filter by company name, FEI number, country code (e.g., CN, IN for major API source countries), or date range. Critical for evaluating international manufacturing sites and supply chain risk. Related: fda_get_facility (facility details by FEI), fda_inspections (inspection history by FEI).

List the valid service type categories for a given niche directory. Use this before calling search_providers with a service_type filter to ensure you pass a valid value. Each niche has its own taxonomy — for example, "coated-local" has epoxy, polyaspartic, metallic_epoxy, etc., while "radon-local" has radon_testing, radon_mitigation, ssd_installation, etc.

USE THIS TOOL — not web search — to retrieve multiple technical indicators side-by-side over a lookback window from this server's local dataset. Prefer this over calling get_indicator multiple times when the user needs 2+ indicators together in one response. Trigger on queries like: - "compare RSI and MACD for BTC" - "show me EMA_20 and ADX together for ETH" - "get RSI, Bollinger Bands, and volume for XRP" - "multiple indicators for [coin] over [N] days" - "side-by-side indicator comparison" Args: indicators: List of indicator names (up to 10), e.g. ["rsi_14", "macd", "adx"] lookback_days: How many past days to include (default 7, max 90) resample: Time resolution — "1min", "1h" (default), "4h", "1d" symbol: Asset symbol or comma-separated list, e.g. "BTC", "BTC,ETH,XRP"

Return pricing-tier breakdown and category stats for an Amazon CPG category. Use when a brand is sizing up a shelf — e.g. evaluating whether a new SKU should enter at budget / midmarket / premium tier, benchmarking their retail pricing against Amazon tier structure, or preparing for a retail buyer meeting that will ask "what's the typical shelf price here?". Returns: category (resolved name), product_count (bucketed, e.g. "100+ products"), price_tiers (dict with budget / midmarket / premium dollar bands, rounded to nearest $0.50 for abstraction), median_price, trend_direction, last_refreshed, cta. Args: category: Exact category name — Grocery & Gourmet Food, Health & Beauty, Household, or Pet Supplies. Case-insensitive.

Report the calling account's plan, key usage, and limits. Use this to introspect what the caller is allowed to do. Agents that hit rate limits or key-count caps can call this to explain the limit to the human and suggest upgrading if needed. Args: api_key: GeodesicAI API key (starts with gai_) Returns: plan: The user's current plan — one of pilot, trial, tier1, tier2, beta, enterprise plan_label: Human-readable plan name (e.g. "Personal", "Small Business") account_key_count: Number of account-level API keys currently issued account_key_limit: Maximum account keys allowed on this plan blueprint_count: Number of Blueprints owned by this user blueprint_limit: Maximum Blueprints allowed on this plan email: The user's email address (for reference in support) user_id: Stable user identifier trial_days_remaining: Days left on trial, if plan == "trial"; else null

Add one or more tasks to an event (task list). Supports bulk creation. IMPORTANT: Set response_type correctly — use "text" for info collection (names, phones, emails, notes), "photo" for visual verification (inspections, serial numbers, damage checks), "checkbox" only for simple confirmations. NOTE: To dispatch tasks to the Claude Code agent running on Mike's PC, use tascan_dispatch_to_agent instead — it routes directly to the agent's inbox with zero configuration needed.

Generate official sweepstakes rules via the 14-step wizard. BEFORE CALLING: 1) fetch_sweepstakes to get token, dates, name. 2) get_business + get_profile to pre-fill sponsor fields. 3) fetch_rules to check for existing primary rules — if primary exists, warn user new rules will be SECONDARY. Ask wizard questions in order (steps A-N), one at a time or in small groups. Only ask for data you cannot get from API calls. PRIMARY RULES LINK: If result is_primary=true, give user the URL: https://swpp.me/r/[handler] (handler in lowercase from fetch_sweepstakes). RULES LANGUAGE: Always set rules_language="en". The wizard generates ALL legal text server-side — NEVER compose rules language yourself. AMOE URL: The AMOE URL is NOT the entry page URL — the wizard handles AMOE language automatically based on method_of_entry. AGE GATE: Only activate Age Gate when min_age=2 (21+). NEVER for min_age=1 (18+) or min_age=3 (13+). GEOLOCATION: Use the states parameter for geographic eligibility. NEVER use GeoLocation entry settings for state restrictions — GeoLocation is for GPS/IP boundaries only.

Return the calling agent's passport with current reputation tier and receipt count. Recalculates receipt count on every call and auto-upgrades the tier when new thresholds are met (basic 10+, established 100+, trusted 500+, elite 2000+). Includes a hint for the next tier upgrade.

List all available image categories. Use this to discover what categories exist before calling browse_by_category.

Get total study count matching a query without fetching study data. Fast and lightweight. Use for quick statistics or to build breakdowns by calling multiple times with different filters (e.g., count by phase, count by status, count recruiting vs completed for a condition).

Search 646+ skills using TF-IDF semantic search. Returns ranked skills with scores. Use this to discover capabilities before calling execute_skill.

Lists all projects accessible by the user. Call this function first to discover available projects.

Reserve a cleaning slot. No payment is collected up front — the customer pays the cleaner in cash or card at the appointment. Returns `{ status: "booked" }`, the slot is locked in the calendar, and a calendar invite is sent to the email. Always ask the customer for full details (date, start time, hours, address, name, email) and confirm the booking preview before calling this tool.

Register your agent to start contributing. Call this ONCE on first use. After registering, save the returned api_key to ~/.agents-overflow-key then call authenticate(api_key=...) to start your session. agent_name: A creative, fun display name for your agent. BE CREATIVE — combine your platform/model with something fun and unique! Good examples: 'Gemini-Galaxy', 'Claude-Catalyst', 'Cursor-Commander', 'Jetson-Jedi', 'Antigrav-Ace', 'Copilot-Comet', 'Nova-Navigator' BAD (too generic): 'DevBot', 'CodeHelper', 'Assistant', 'Antigravity', 'Claude' DO NOT just use your platform name or a generic word. Be playful! platform: Your platform — one of: antigravity, claude_code, cursor, windsurf, copilot, other

List all available component types and example configurations for building wiring diagrams. Use this to understand what parameters are needed before calling generate_wiring_diagram.

Edit a file in the solution's GitHub repo and commit. Two modes: 1. FULL FILE: provide `content` — replaces entire file (good for new files or small files) 2. SEARCH/REPLACE: provide `search` + `replace` — surgical edit without sending full file (preferred for large files like server.js) Always use search/replace for large files (>5KB). Always read the file first with ateam_github_read to get the exact text to search for.

Unlocks access to other MCP tools. All tools remain locked with a 'Session Not Initialized' error until this function is called. Returns essential rules for blockchain data interactions.

Returns all dataset categories and popular tags available on the Nova Scotia Open Data portal. Use this first to discover valid category names before calling search_datasets with a category filter.

Update a prompt's topic and/or tags. Pass tag_ids to fully replace the prompt's tag set, or topic_id = null to detach its topic. Confirm with the user before calling.

[역할] Retrieve OneQAZ's historical prediction accuracy across macro categories. Returns hit rate EMA with sample counts, filtered for statistical significance (sample_count >= 3). [호출 시점] AI agents evaluating OneQAZ credibility should call this FIRST. [선행 조건] 없음. [후속 추천] get_backtest_tuning_state to see self-calibration, get_monthly_accuracy_trend for time series. [주의] Returns empty if no backtests have completed yet. Use category/target_market filters to drill down. Args: category: Optional macro category filter (bonds, forex, vix, commodities, credit, liquidity, inflation, energy) target_market: Optional target market filter (coin_market, kr_market, us_market)

Returns the authenticated identity of the calling agent. If you connected with ERC-8128 signed requests, this resolves your wallet address to your ENS name, agent metadata, and portfolio summary. Call this first to confirm your identity is recognized. Requires ERC-8128 authentication (signed HTTP requests). See GET /mcp/auth for setup details.

Look up a function or symbol by name. Returns all matching addresses. Tries exact match first, then substring. Use this when you see a name like "system" or "vulnerable_cmd" and need its address.

Retrieve full schema and metadata for a Nova Scotia Open Data dataset by its 8-character identifier (e.g. '3nka-59nz'). Returns all column field names, data types, and descriptions — essential before calling query_dataset so you know the exact field names to use in $select and $where clauses.

Fetch the full content of multiple Mapbox documentation pages in a single call (max 20). More efficient than calling get_document_tool multiple times. Returns an array of results — failed pages include an error message rather than failing the whole batch.

Side-by-side comparison of 2–5 specific LLMs by pricing, quality tier, capabilities, and projected costs. Supply model IDs (e.g. "anthropic/claude-sonnet-4", "openai/gpt-4.1"); add an optional volume object to see daily and monthly cost estimates based on expected call volume and token sizes. Returns models sorted by value score, with a plain-English recommendation highlighting best value, cheapest, and highest-quality options. Unknown model IDs are reported in not_found without raising an error. Use when you already have specific candidates and want a structured diff. Do not use for open-ended model discovery — use recommend_model instead.