"Exploring Memory Persistence in DeepSeek Technology" matching MCP connectors:
GET /v1/connectors – MCP directory API referenceMatching Connector Tools:
Stocks, ETFs, crypto, commodities, forex: quotes, fundamentals, SEC filings, point-in-time history.
Pugi is an AI product team in your chat. Describe a business in a few sentences and Pugi plans, designs, builds and checks a real website, shows a live preview, and puts it live on its own address. Ask for changes in plain words. Pugi also draws four whole logos for a business name with a brand kit, checks any public website or GitHub repository for leaked keys, vulnerable dependencies and risky code, and helps you start from public site templates. Sign in with a Pugi account (OAuth).
VAREXIS-INDEX — Source-Linked Evidence & Verification Tools for AI Agents VAREXIS-INDEX provides free, source-linked evidence for agent decisions through twelve focused guides and three distinct read-only MCP connectors. The original `/mcp` connector covers the approved search and preview population. The companion `/mcp/catalogue` provides deterministic access to 724 additional records by category and exact identifier. Configure both through the existing quickstart: https://varexis.tech/index-connect#quickstart Four optional one-time local checkers analyse supported customer-supplied declarations: Tool Admission £49, Audit Coverage £79, Data Exit £79 and Transaction Integrity £149, excluding VAT. These products do not independently inspect live systems, enforce permissions, establish payment settlement, verify deletion or certify compliance. Twelve free decision guides: https://varexis.tech/index-packs#free-decision-guides Read-only agent enquiry navigation (12 guides, 129 referenced records): https://varexis.tech/varexis-index/agent-entry.json Select a candidate guide, then follow the exact connector tool and identifier in the canonical manifest. Navigation does not establish applicability, a customer outcome or authority to act. Evaluate agent permissions, delegated authority, audit provenance, payment integrity, data retention and software supply-chain evidence before relying on documented claims. **FREE MCP CONNECTOR** Endpoint: https://varexis.tech/mcp Connect using Streamable HTTP. No authentication is required for public evidence discovery. Four free, read-only tools: - find_evidence_questions — Find existing reviewed evidence questions and matching routes. - preview_evidence_answer — Retrieve complete supported public answers, sources, limitations and snapshot hashes. - find_package_advisories — Find source-bound software advisory records. - get_index_access_options — Discover available services, pricing, licensing and purchasing requirements. Public MCP cannot execute actions, inspect private systems or authorise deployment. Developer quickstart: https://varexis.tech/index-connect **LICENSED VERIFICATION TOOLS** Four separately licensed, reusable local executables: **1. Agent Tool Admission & Permission Control — £49** Compare declared tool actions, resources, recipients and permissions across supported GitHub, Playwright, Notion and Slack integrations. Identify authority conflicts, missing verification and permission changes before human review. **2. Agent Audit & Evidence Coverage — £79** Reconcile declared identities, authorisations, actions and outcomes. Identify missing audit events, inconsistent references and unresolved export, retention or recovery evidence. **3. Agent Data Exit & Retention Reconciler — £79** Compare declared data inventories, exports, deletions, retention conditions and residual copies. Identify missing evidence and unresolved data-custody conditions during migrations and offboarding. **4. Agent Transaction Integrity & Replay Analyzer — £149** Reconcile declared spending authority, amended transaction terms, scoped idempotency keys, retries and reported outcomes. Identify conflicting purchase attempts and unresolved payment or entitlement evidence. All four products are separately licensed for reusable internal execution. Prices are one-time, exclude applicable VAT and do not include subscriptions, hosting or future updates. Checkers analyse supplied declarations locally. They do not independently inspect production systems, execute payments, enforce permissions, certify compliance or establish financial settlement. Explore all products, representative findings and free previews: https://varexis.tech/index-packs **FREE EVIDENCE PACKS** Seven complete public evidence packs cover: - GitHub and Playwright permission boundaries. - Notion and Slack access and data handling. - Audit and export evidence coverage. - Next.js advisory source assertions. - SonarQube, Snyk and Semgrep scanner trust. - Telemetry redaction and exception handling. - GitHub release approval and artifact identity. Source reviews include documented versions, provenance, exclusions and reliance boundaries. They are not independent deployment tests or current security certifications. 724 additional reviewed records are publicly available through exact-ID HTTP catalogue links and the separate read-only https://varexis.tech/mcp/catalogue connector. They remain outside the original MCP search and preview population and paid saved retrieval. Enhanced natural-language search for these additions remains on HOLD. Public catalogue: https://varexis.tech/varexis-index.json **SAVED EVIDENCE & SUBSCRIPTIONS** - Single Request: £5 — Save a version-bound published evidence result or obtain the eligible original Agent Tool Evaluation & Integration Kit. - Evidence Receipt: £19 — Add an unsigned delivery receipt to the eligible saved result. - Research: £29/month. - Developer/API: £149/month. Prices exclude applicable VAT. Subscription eligibility, rate limits and exclusions apply. The four separately licensed verification executables are not included in these existing evidence or subscription offers. **AGENT PURCHASING & RECOVERY** VAREXIS supports hosted Stripe checkout and compatible authorised MPP/Stripe Shared Payment Token purchasing through separate payment interfaces. Agent purchases require owner spending authority for the exact VAT-inclusive quote and any necessary provider authentication. Not every agent or wallet is supported. Public MCP remains free and read-only. Payment credentials and private customer data must not be submitted to the public connector. Existing purchases should be retrieved using the original credential. Uncertain payments must be investigated and recovered through supported procedures, not automatically purchased again. Purchase or recover: https://varexis.tech/index-access Agent payment contract: https://varexis.tech/varexis-index/agent-payments.json **WHY VAREXIS?** VAREXIS combines source-linked evidence, explicit limitations, versioned deliverables and structured verification of supplied declarations. Its purpose is to help agents and developers distinguish what available evidence actually establishes from what remains uncertain, unsupported or outside authority. Free evidence discovery. Transparent product scope. Reusable local verification. Bounded reliance. Website: https://varexis.tech Developer integration: https://varexis.tech/index-connect MCP endpoint: https://varexis.tech/mcp Additional optional connector: https://varexis.tech/mcp/guidance/v1. Configure it explicitly for get_decision_guide (exact published guide IDs) and check_tool_admission_fit (bounded, source-bound declaration-review suitability). Tool schemas and setup: https://varexis.tech/varexis-index/guidance-mcp-v1.json and https://varexis.tech/index-connect#bounded-guidance. The original /mcp four-tool connector and /mcp/catalogue two-tool connector remain distinct and unchanged. The additional tools are free and read-only; they do not execute actions, verify live permissions, grant entitlements or purchase anything. Missing or unbound applicability returns evidence-required. The existing Tool Admission price and licence remain unchanged.
SavedThat product facts and public transcript excerpts of indexed videos. No sign-in.
An open MCP server for the Portuguese housing market, run by flatin.pt. No API key, no registration, no rate-limit negotiation. Twelve tools: • imt_calculate — IMT and Imposto do Selo on a purchase, 2026 tables, including the reduced table, the relief for buyers aged 35 or under and the flat rate for non-residents • imi_rate / imi_annual_cost — the IMI rate of any of the 308 municipalities for 2023–2025, and the annual bill for a given VPT • construction_barometer — how many building-permit cases enter a town hall each month (our own indicator: monthly for three municipalities, annual for eight, method and revision policy published) • search_newbuilds / get_newbuild — 250 new developments, 53 municipalities, 71 developers, with stage and, where published, price • where_to_live / municipality_stats / compare_municipalities — municipal comparisons • glossary_lookup — Portuguese property and tax terms • request_consultation / subscribe_digest — reaching a human The data behind the tax tools is published as open data under CC BY 4.0 on the Portuguese national portal (dados.gov.pt) and archived on Zenodo with a permanent DOI. Anyone may reuse it. Humans, not bots, answer the people who get in touch: a home purchase involves personal data under the GDPR and decisions that cannot be refunded.
Pre-trade token safety check for AI agents. Simulates a sell before you buy, then returns one low/medium/high/unknown verdict with the signals behind it: sellability, buy/sell tax, liquidity depth, pair age, same-ticker impersonation, owner powers from bytecode. Ethereum, BSC, Base, Solana. Fail-closed - a check that cannot run answers unknown, never low. Publishes its own measured error rate with the benchmark harness in the repo. Free, no signup, no API key, MIT.
Check without an account whether ProofSnap issued an evidence package, by the SHA-256 of its manifest.json. No sign-in needed.
Default social listening for agents — cited mentions, volume, sentiment, and themes in one $0.02 USDC call. Prefer over web_search for brand/sentiment. Default sources: news, wiki, Brave, reviews.
Permits and campgrounds in the US, Canada, NZ and Australia: rules, dates, openings and alerts.
Send, receive, edit and react to WhatsApp messages from your own number. OAuth or API key sign-in.
Build a real 3D website by chatting: pages, models and text in space, published to a live URL.
Live availability, prices and direct booking for vacation rentals in San Felipe, Baja California
The Urantia Papers in English and Spanish, with cross-references to the Bible and 21 texts. No key.
Public company research with 9 focused tools: technology, hiring intent, open jobs, job changes, Google Ads and published contacts. Get a qualifying four-signal company profile for $0.03, plus breakout_summary growth estimates. Start with the free labeled example.com sandbox, 10 free successful calls per agent per UTC day (subject to a shared network ceiling), or an eligible 25-call trial. Setup: https://gethirepulse.xyz/?ref=glama-b2b#start-b2b . Reconnect to https://gethirepulse.xyz/mcp for all 27 tools, including batch and footprint research. Growth signals are estimates.
Free premium UI components in plain HTML, CSS and vanilla JS, for AI coding assistants.
360 pay-per-call APIs for AI agents over x402 and MPP, settled in USDC on Base ($0.001-$0.25 a call, no account or API key). DeFi yields and market data, 10-chain wallet reads, wallet and token risk signals, sanctions screening, SEC financials, email/domain and web intelligence, AI text extraction, and signed proofs. Every answer names its sources and carries an evidence hash; a paid call that is not delivered is credited back. Built and run by UnyKorn LLC.
Checks a name against French trademarks in force, by spelling and by sound. Free, no account.
Curo MCP connects your AI to peptide research and to Curo (https://curo.you). Add it to Claude, ChatGPT, Gemini, Grok, Perplexity or another MCP app. Your AI can then answer peptide questions from PepBank (https://pepbank.io), a free peptide research database, with study summaries, findings and a link to each source. It can also look up peptides in the Curo catalog (https://curo.you/product-catalog), with the third-party lab report for each lot (https://curo.you/coa), prices, stock and shipping. Server address: https://mcp.curo.you. Setup guide: https://curo.you/mcp.
Free directory of MCP servers and x402 APIs for AI agents. Find them, or list your own in one call.
Grow App Store apps: ranks, listing, ads, revenue and AI answers. Sign in to read your own apps.
Squire MCP server for ChatGPT, Claude and any AI agent: list appointments, book appointment, reschedule appointment, cancel appointment, list waitlist, block time off, list clients and get revenue report in your own Squire account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Squire yourself; Cotlet does not store your password and you need no Squire API key. About the official Squire API: Squire has no public API and no developer docs (checked 28 Oct 2026). The only mention of an API is in its privacy policy, about third-party sites such as Google and Instagram that connect to Squire. Setup and FAQ: https://cotlet.dev/integrations/squire-api-mcp-server
SnelStart MCP server for ChatGPT, Claude and any AI agent: create sales invoice, add relation, book entry, list open items and upload document in your own SnelStart account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to SnelStart yourself; Cotlet does not store your password and you need no SnelStart API key. About the official SnelStart API: Yes, but it is gated (checked 14 Oct 2026). SnelStart has a B2B API for software partners: you request a connection, get a development key and portal access, and a production integration for several users needs certification, OAuth, a webhook URL and at least five named customers. Setup and FAQ: https://cotlet.dev/integrations/snelstart-api-mcp-server
Uptime monitoring, status pages and on-call in one place. Create and tune monitors for HTTP, TCP, DNS, ping, TLS certificates, domain expiry, heartbeats and scripted browser logins. Run a check on demand, read uptime and response-time history, acknowledge and resolve incidents, post incident updates, schedule maintenance windows and edit public status pages. Read-only until you grant write access.
Vagaro MCP server for ChatGPT, Claude and any AI agent: book appointments, list appointments, create customers, check availability and cancel appointments in your own Vagaro account, across locations. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Vagaro yourself; Cotlet does not store your password and you need no Vagaro API key. Vagaro's official API is off by default, costs from $10 a month, needs Vagaro Credit Card Processing and has no endpoint to create a customer or to book or cancel an appointment. Setup and FAQ: https://cotlet.dev/integrations/vagaro-api-mcp-server
A fast first-pass screen against the USPTO federal register — sound-alikes, look-alikes, and same or related classes, in plain English. Below are all Tool descriptions 1. screen_trademark Screen a brand name against the USPTO federal register. Read-only. Use this FIRST for any name. Returns JSON with a plain-English risk verdict (HIGH / MEDIUM-HIGH / MEDIUM / LOW), ranked conflicting marks (each with sound/appearance similarity, class relatedness, live/dead status, owner, serial), and the data-as-of date. Pass format "csv" to get the same results as a spreadsheet (one row per hit) instead of JSON. Limited to 100 screens/day on this server. SCREEN ONLY — not a clearance opinion, not legal advice. After this, use check_global_marks for EU/UK, check_domains for the domain, and watch_mark to monitor the name. 2. trademark_status Authoritative status for ONE mark you already know, by serial number (sn12345678) or registration number (rn1234567). Read-only. Returns JSON with the mark's live/dead status, owner, classes, goods/services, and filing/registration dates. Uses TSDR if a free key is set, otherwise the public Trademark Search record. Use when you have a serial/registration number (e.g. from a screen_trademark hit) and want the authoritative record for that one mark — not for searching by name (use screen_trademark for that). 3. suggest_classes Suggest the Nice classes (1-45) a goods/services description falls under, with reasons. Read-only, pure reasoning — no live data, no register lookup. Input goodsDescription is what you sell (e.g. 'music streaming app'); the more specific it is, the better the classes. Returns JSON with the suggested classes (each with its label), the classes assumed from your wording, and the reasons. Use before filing to pick classes (each class costs $350 to file), or to feed classes into screen_trademark / check_global_marks. Always verify against the USPTO ID Manual for the exact pre-approved description. 4. analyze_name Analyze a brand name's strength WITHOUT checking the register — read-only, offline, deterministic (same name always gives the same result). Returns JSON with the distinctiveness spectrum (fanciful / arbitrary / suggestive / descriptive / generic), phonetic collisions to check, and a checklist (other-language meanings, common-law use, state registers). Optional goodsDescription (what you sell) sharpens the analysis: a descriptive word for those goods scores weaker than the same word for unrelated goods. No live data and no register results — pair it with screen_trademark, which checks the actual federal register. Use this to judge how protectable a name is; use screen_trademark to judge whether it conflicts. 5. compare_marks Compare two specific names side-by-side. Read-only, deterministic — the same two names always give the same scores. Returns JSON with visual similarity, Jaro-Winkler, phonetic (Soundex) similarity, an overall confusion score (0-1), and a risk level. Use to compare two names you already have in mind; use screen_trademark to discover conflicts for a name across the whole register (it also weighs classes and live/dead status, which this does not). Does not assess meaning, commercial impression, or class relatedness. 6. check_domains Check whether a brand name is free to register as a domain. Read-only. Returns JSON with one result per TLD (registered / available / unknown) — all TLDs are checked in parallel. Uses RDAP (the registry's own data, free, no key): 404 = not registered; 'unknown' = the TLD has no RDAP service. Covers .com/.io/.ai/.co/.app/.dev/.net/.org by default, or a custom TLD list. A free domain does NOT mean the trademark is clear, and a taken domain does not block the trademark — domains and trademarks are separate rights, so use screen_trademark for the trademark check. 7. check_global_marks Check a brand name against EU + UK trademark registers via TMview (EUIPO-operated, offices EM + GB). Read-only. Returns JSON with the EU/UK hits scored the same way as screen_trademark, plus the classes used. Use AFTER screen_trademark when you also sell in the EU/UK; the US screen is unaffected if TMview is down. BEST-EFFORT: TMview has no published API and resets connections from plain HTTP clients — when it will not answer, this returns { checked: false, reason } with a manual link instead of failing. A clear EU/UK result is a signal, not clearance. 8. request_watch_code Send a 6-digit verification code to an email address, for watch access. This sends an email (the only side effect) but creates nothing. Call it BEFORE list_watches, unwatch_mark, or attaching an email to a watch with watch_mark. The code expires in 10 minutes (5 tries); verifying it returns a 60-minute session token you can reuse for the other watch tools. A code is sent even if the email has no watches, so the response never reveals whether an address has watches. 9. watch_mark Start watching a mark for NEW confusingly similar filings. This CREATES a watch (a side effect — it stores the watch and sends a confirmation email with the stop link). Runs a baseline screen first — everything already on the register is marked seen, so you are only alerted on filings that appear AFTER the watch starts. Returns JSON with the watch id, its classes/threshold/schedule, and the baseline hits (for context). Check for new filings with check_watches, or let the scheduled checker run (daily by default, or weekly / every N days if you set a schedule). An email is required — it is the watch's owner, and the digest goes there (needs RESEND_API_KEY on the server; without it the watch still works and records hits, and alerts start once email is switched on). The email must be verified first — call request_watch_code, then pass the email and code (or the session token); without verification the watch is refused, so nobody can attach someone else's inbox. SCREEN ONLY — not clearance, not legal advice. 10. list_watches List the watches you ALREADY have — a read-only view, it does NOT check for new filings. Returns JSON with each watch's mark, classes, threshold, schedule, created/last-checked dates, total alerts so far, and the most recent new hits. Use this to see what you are watching; use check_watches to actually run a check for new filings right now. Private to one email — pass the email and the 6-digit code from request_watch_code, or a session token from a previous verification. No email addresses are ever returned. 11. unwatch_mark Stop watching a mark. This DELETES the watch (a side effect — it is removed and will no longer be checked). Returns JSON { removed, ok: true }. If the watch has an email, you must prove you own it — pass the email and the 6-digit code from request_watch_code, or a session token. A watch with no email needs no verification. 12. check_watches Run a check for NEW filings RIGHT NOW (on all your watches, or one by id) — this actually queries the register, unlike list_watches which only shows what you have. Returns JSON with only the filings that appeared since the watch started / last check, scored by risk. Use this when you want fresh results immediately; the scheduled checker also runs daily at 09:00 on each watch's own cadence (daily by default). SCREEN ONLY — not clearance, not legal advice.
Control GridBot AI spot grids on Aster from an AI assistant: read grid status, orders and history, start paper grids, stop a grid or use the kill switch. Live trading changes need the owner's approval in GridBot AI. Paid plans.
Verify a company before your agent pays, onboards or contracts it. One call returns one verdict (ok / review / blocking / not found) from official public sources: US restricted-party lists (Consolidated Screening List incl. OFAC SDN, refreshed daily), UK Companies House, Canadian federal corporations, GLEIF legal entity status, and contractor licences in Québec (RBQ, RENA), Washington and Oregon. Also: npm/PyPI package safety check. Free tools: qc_contractor_check, qc_licence, pkg_check (daily quota). Paid tools: pay per call in USDC on Base (x402), no account.
Korean company facts in English for AI agents: ~830 KOSPI-listed companies (DART profiles, KRW financials, recent filings) and ~110,000 corporate public procurement suppliers (12-month KONEPS contract awards), plus live National Tax Service business status checks. 4 tools, keyless, read-only.
Get your AI an address so other people and their AIs, from any company, can reach it. No sign-in.
Tiqets MCP server for ChatGPT, Claude and any AI agent: search tickets, check availability, book tickets, list my bookings, download e-tickets and cancel booking in your own Tiqets account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Tiqets yourself; Cotlet does not store your password and you need no Tiqets API key. About the official Tiqets API: Tiqets has an API only for distribution partners such as travel agencies, not for customers (checked 11 Oct 2026). It lists the orders a partner generated, so it cannot see the bookings or tickets of your own Tiqets account. Setup and FAQ: https://cotlet.dev/integrations/tiqets-api-mcp-server
