Browse 159,125 AI agents

By Sm0k13

B.R.O-abliterated-27B is an open-weight text-generation model with roughly 27B parameters published by Sm0k13 on the Hugging Face Hub. It has 767 downloads. Runs with mlx.

By MissingPackage

Qwen3.5-0.8B-knapquant-GGUF is an open-weight text-generation model with roughly 0.8B parameters published by MissingPackage on the Hugging Face Hub. It has 688 downloads. Runs with gguf.

By el4

Sage-A1.5-TOFFEE-GGUF is an open-weight text-generation model published by el4 on the Hugging Face Hub. It has 882 downloads.

By kybrcore

Hemmingway-1-NInfer is an open-weight text-generation model published by kybrcore on the Hugging Face Hub. It has 810 downloads. Runs with ninfer.

By SupraLabs

Supra2.1-Ultra is an open-weight text-generation model published by SupraLabs on the Hugging Face Hub. It has 588 downloads. Runs with transformers.

By Verdugie

Candid-3.5 is an open-weight text-generation model published by Verdugie on the Hugging Face Hub. It has 1,906 downloads. Runs with gguf.

By mradermacher

Qwen3.5-9b-Opus-Openclaw-Distilled-GGUF is an open-weight text-generation model with roughly 9B parameters published by mradermacher on the Hugging Face Hub. It has 900 downloads. Runs with transformers.

By dongchengli

gemma-3-4b-it-text-int4-gptq is an open-weight text-generation model with roughly 4B parameters published by dongchengli on the Hugging Face Hub. It has 2,266 downloads. Runs with transformers.

By Verdugie

Fable-Therapy is an open-weight text-generation model published by Verdugie on the Hugging Face Hub. It has 1,015 downloads. Runs with gguf.

By Feyd89

Qwen3.8-Flash-Next-NVFP4-capy-flash is an open-weight text-generation model published by Feyd89 on the Hugging Face Hub. It has 1,082 downloads. Runs with ninfer.

By AMAImedia

Naive-N0.5-Flash-Mixed-Quant-GGUF is an open-weight text-generation model published by AMAImedia on the Hugging Face Hub. It has 1,148 downloads. Runs with gguf.

By prithivMLmods

Sage-A1.5-MXFP4_MOE-MTP-GGUF is an open-weight text-generation model published by prithivMLmods on the Hugging Face Hub. It has 1,630 downloads. Runs with transformers.

By nchapman

Qwen3.5-4B-DFlash2 is an open-weight text-generation model with roughly 4B parameters published by nchapman on the Hugging Face Hub. It has 1,226 downloads. Runs with speculators.

By darioooooo0o

Qwen3.6-35B-A3B-TinyKV-GGUF is an open-weight text-generation model with roughly 35B parameters published by darioooooo0o on the Hugging Face Hub. It has 1,301 downloads.

By Feyd89

Qwen3.8-Flash-Next-GSQ-RCO-capy-flash is an open-weight text-generation model published by Feyd89 on the Hugging Face Hub. It has 1,184 downloads. Runs with ninfer.

By ibm-granite

granite-4.2-3b-fp8 is an open-weight text-generation model with roughly 3B parameters published by ibm-granite on the Hugging Face Hub. It has 1,433 downloads. Runs with transformers.

By ryugyosoft

granite-4.2-3b-onw is an open-weight text-generation model with roughly 3B parameters published by ryugyosoft on the Hugging Face Hub. It has 1,416 downloads. Runs with onw.

By dealignai

Nanbeige4.3-33B-UNCENSORED-GGUF is an open-weight text-generation model with roughly 33B parameters published by dealignai on the Hugging Face Hub. It has 9,874 downloads.

By ReadyArt

Wicked-Tutu-26B-A4B-Final-GGUF is an open-weight text-generation model with roughly 26B parameters published by ReadyArt on the Hugging Face Hub. It has 1,829 downloads.

By lflare

Qwen3.8-Flash-Next-NVFP4-Abliterated is an open-weight text-generation model published by lflare on the Hugging Face Hub. It has 1,651 downloads. Runs with transformers.

By ReadyArt

Omega-Arsenal-12B-GGUF is an open-weight text-generation model with roughly 12B parameters published by ReadyArt on the Hugging Face Hub. It has 10,608 downloads.

By QuantFactory

stable-code-instruct-3b-GGUF is an open-weight text-generation model with roughly 3B parameters published by QuantFactory on the Hugging Face Hub. It has 2,427 downloads. Runs with transformers.

By jackasda211233

Qwen3.5-27B-Uncensored-RYS-Reasoner-GGUF is an open-weight text-generation model with roughly 27B parameters published by jackasda211233 on the Hugging Face Hub. It has 2,221 downloads.

By Nanbeige

Nanbeige4.3-33B is an open-weight text-generation model with roughly 33B parameters published by Nanbeige on the Hugging Face Hub. It has 4,785 downloads. Runs with transformers.

By abenzerps

Nanbeige4.3-33B-GSQ-RCO-GGUF is an open-weight text-generation model with roughly 33B parameters published by abenzerps on the Hugging Face Hub. It has 8,606 downloads. Runs with gguf.

By PollardWeights

NVIDIA-Nemotron-3.5-Bolt-E4B-Pollard is an open-weight text-generation model with roughly 4B parameters published by PollardWeights on the Hugging Face Hub. It has 5,348 downloads. Runs with gguf.

By nesymerp1

Nanbeige4.3-33B-Base-Q4_K_M-GGUF is an open-weight text-generation model with roughly 33B parameters published by nesymerp1 on the Hugging Face Hub. It has 3,676 downloads. Runs with transformers.

By bartowski

PocketDoc_Dans-PersonalityEngine-V1.2.0-24b-GGUF is an open-weight text-generation model with roughly 24B parameters published by bartowski on the Hugging Face Hub. It has 5,167 downloads.

By haoranxu

Llama-3-Instruct-8B-SimPO is an open-weight text-generation model with roughly 8B parameters published by haoranxu on the Hugging Face Hub. It has 8,587 downloads. Runs with transformers.

By sbintuitions

sarashina2.2-3b-instruct-v0.1 is an open-weight text-generation model with roughly 3B parameters published by sbintuitions on the Hugging Face Hub. It has 13,391 downloads.

By mobilint

EXAONE-4.0-1.2B is an open-weight text-generation model with roughly 1.2B parameters published by mobilint on the Hugging Face Hub. It has 27,406 downloads. Runs with transformers.

By trl-internal-testing

tiny-Phi3ForCausalLM-3.5 is an open-weight text-generation model published by trl-internal-testing on the Hugging Face Hub. It has 23,785 downloads. Runs with transformers.

By bartowski

Nanbeige_Nanbeige4.3-33B-GGUF is an open-weight text-generation model with roughly 33B parameters published by bartowski on the Hugging Face Hub. It has 62,465 downloads.

By ibm-granite

granite-4.2-30b is an open-weight text-generation model with roughly 30B parameters published by ibm-granite on the Hugging Face Hub. It has 40,121 downloads. Runs with transformers.

By mikeinnyc

Qwen3.8-27B-GPTQ-Int4-sym-G128-MTP-BF16 is an open-weight text-generation model with roughly 27B parameters published by mikeinnyc on the Hugging Face Hub. It has 34,387 downloads. Runs with transformers.

By patrickbdevaney

GLM-5.3-Flash-REAP50-GGUF is an open-weight text-generation model published by patrickbdevaney on the Hugging Face Hub. It has 33,177 downloads. Runs with gguf.

By AtomicChat

Qwen3.5-4B-DFlash-GGUF is an open-weight text-generation model with roughly 4B parameters published by AtomicChat on the Hugging Face Hub. It has 28,956 downloads. Runs with gguf.

By FengZhenfei

Ontology-Augmented Generation for AI agents. An open-source AI framework/SDK by FengZhenfei, hosted on GitHub.

By busabase

Open-source database & workspace for AI agents — structured data, durable knowledge, reusable skills, runnable apps, and human review on the writes that matter. Local-first and self-hostable. An open-source AI framework/SDK by busabase, hosted on GitHub.

By dagoSte

Coordinate Claude, Codex and Cursor windows with shared tasks, acknowledgements, prechecks, parallel tests, source-bound proofs and independent review. Built-in skills recall lessons, simplify code and forecast work. Track progress, decisions and evidence in a bilingual local dashboard, with MCP plugins. Browse skills and verified security audits. An open-source AI framework/SDK by dagoSte, hosted on GitHub.

By karust

Self-hosted web search for AI agents and SEO. Google, Bing, Yandex, Baidu, DuckDuckGo & Ecosia, rendered in a real browser: ranked results and clean page Markdown. API, CLI, MCP 🎉. An open-source AI framework/SDK by karust, hosted on GitHub.

By michaelshimeles

Agent skills and an AGENTS.md workflow template — isolate in worktrees, build to a service layer, prove with evidence, ship with before/after proof and Greptile review loops. For Claude Code, Cursor, and Codex. An open-source AI framework/SDK by michaelshimeles, hosted on GitHub.

By mmalmi

Private mesh VPN for humans and agents. Local keys, signed enrollment, no provider account. Native apps and CLI. An open-source AI framework/SDK by mmalmi, hosted on GitHub.

By TYKAIRO-AI

Quick overview This webhook-driven workflow generates short-form video scripts and publishing copy with OpenAI, runs a local copy/duration audit, and renders approved 9:16 videos via Creatomate, returning either a creative kit, an audit checklist, a render job ID, or a final video URL. How it works Receives a POST webhook request (header-authenticated) with an action of draft, audit, render, or status. Validates and sanitizes the request fields, enforcing limits and requiring explicit approval and rights confirmation before rendering. For draft requests, sends a prompt to the OpenAI Chat Completions API to generate two hooks, a 4-scene storyboard, voiceover, hashtags, and SEO metadata, then returns a validated publishing kit for YouTube/TikTok/Instagram. For audit requests, performs a deterministic local preflight to estimate duration and flag copy issues (for example length limits and exaggerated claims) and returns a checklist for human review. For render requests, submits the approved Hook, Supporting_Text, and CTA into a specified Creatomate 1080×1920 template and returns the queued render job ID. For status requests, fetches the Creatomate render job status and returns the final video URL and snapshot URL once the render succeeds (or errors if it fails). Setup Create a header-authenticated webhook credential for the incoming request and configure the calling app to send the required header. Add an OpenAI API key as an HTTP Header Auth credential for the OpenAI request step. Add a Creatomate API key as an HTTP Header Auth credential for the Creatomate render and status request steps. In Creatomate, create a 1080×1920 template and ensure it contains elements named exactly Hook, Supporting_Text, and CTA, then use its template_id in render requests. Build your client request payloads for each action (draft/audit/render/status) and only call render with approved=true and rights_confirmed=true after a manual review. An n8n automation workflow template by TYKAIRO-AI.

By toshiyuki kikuchi

Quick overview Audits only changed Figma frames against your design system: contrast, font size, tap targets, unlinked tokens and detached components. A Gemini agent reviews each screenshot with your design guide. New issues become Figma comments, fixed ones get a ✅ reply, and Slack gets summaries and weekly reports. How it works Triggers when a Figma named version is saved via webhook, when someone submits the audit request form, or every Monday at 9:00. Validates the request (including the Figma webhook passcode), loads the design guide from Google Docs, and fetches the Figma file structure. Selects the target page, lists top-level frames, compares frame fingerprints against the stored snapshots in n8n Data Tables, and keeps only frames that changed (or forces a review for form requests). For each changed frame, runs deterministic checks on the Figma node tree for contrast, font size, tap targets, token/style usage, and possible detached components, then loads any still-open findings for that frame. Exports a PNG screenshot from Figma and sends it, the layer index, the design guide, and open issues to a Google Gemini agent that proposes new issues and marks previously reported AI issues as resolved when they appear fixed. Reconciles new, still-open, and fixed findings by pinning new comments in Figma, replying in existing threads for fixed items, and updating the findings and frame snapshots in n8n Data Tables. Posts an audit summary to a Slack channel after each run and posts a weekly design-health report to Slack based on the stored findings and frame metrics. Setup Create a Figma personal access token with access to read file content and write comments, and configure it as HTTP Header Auth for all Figma API requests (including the layer-inspection tool) and set a webhook passcode. Add Google Docs OAuth credentials for reading your design guide and update the design guide document URL in the workflow configuration. Add a Google Gemini API credential for the Gemini chat model used by the agent. Add Slack credentials and set the target channel ID in the workflow configuration. Update the workflow configuration values (page name, review focus, thresholds, limits, and timezone), and if using the Figma trigger, register a Figma FILE_VERSION_UPDATE webhook pointing to the production webhook URL with the same passcode. Requirements Figma account with a personal access token (file content read, comments write) Google Gemini API key Google account with the design guide in Google Docs Slack workspace with a bot (chat:write) invited to the target channel n8n with Data Tables (n8n Cloud or a recent self-hosted version) Customization Change contrast, font size and tap-target thresholds or the per-run limits in Workflow Configuration Set reviewLanguage to get Figma comments in another language Add your own rules in Run Design Checks; new rules are tracked and closed automatically like the built-in ones Replace the Slack nodes with Gmail or Microsoft Teams, or move the weekly report to another day Additional info Tested end to end on a Figma file with color variables, text styles, a component and a detached copy. On the second run, fixed issues received a "✅ Fixed" reply and reaction in their original comment threads, and unchanged frames were skipped. An n8n automation workflow template by toshiyuki kikuchi.

By Yuji

Quick Overview This workflow checks the Japan Meteorological Agency (JMA) weather warning feed every 30 minutes, compares changes per watched prefecture, posts alerts and an approval form to Microsoft Teams, and (after approval) emails affected customers via Gmail while tracking shipment and warning state in n8n Data Tables. How it works Runs every 30 minutes and loads workflow settings plus the saved warning state and open shipments from n8n Data Tables. Downloads the public JMA XML warning feed, posts an error card to Microsoft Teams if the feed request fails, and otherwise parses the feed. Filters feed entries to the watched prefectures, keeps the newest entry per prefecture, fetches each entry’s detail XML, and extracts the active warning types and headline. Compares the current warnings to the previously saved warnings per prefecture, routes only newly issued/changed warnings and all-clear events, and upserts the latest state back to the warning state table. For newly issued warnings, matches the affected prefecture (and the warehouse prefecture) against pending shipments due within the lookahead window in the open shipments table. Uses OpenAI to draft a Japanese internal alert plus a polite customer email template with placeholders, then posts the alert to Microsoft Teams and, if shipments are affected, posts a Teams card with a “Review and send” link to an n8n approval form. Waits for staff to approve or skip sending (with a reminder and expiry if no response), then sends one Gmail email per affected order and updates each order record with the send result and posts a send summary to Microsoft Teams. Setup Create two n8n Data Tables named weather_alert_state (prefecture, warnings, reported_at) and open_shipments (order_id, customer_name, email, ship_to_prefecture, ship_date, status, notice_sent, notified_at) and populate open_shipments with your pending orders. Add an OpenAI credential (Chat model) and a Gmail OAuth2 credential for the mailbox that sends customer notices. Configure a Microsoft Teams Workflows/Incoming Webhook URL and paste it into the teams_webhook_url setting, then update watched prefectures, warehouse prefecture, lookahead days, reminder timing, and company signature in the settings node. An n8n automation workflow template by Yuji.

By Onur Olgun

Quick overview This workflow provides a Gemini Live assistant for an e-commerce store by generating single-use Gemini ephemeral tokens, loading a knowledge base from a Google Drive folder, and handling tool calls to read order status, create return/exchange requests, log complaints to Google Sheets and notify the owner via Gmail How it works Receives a POST webhook request to create a new Gemini Live voice session. Applies daily and per-IP rate limits, then lists and downloads supported knowledge files from a specified Google Drive folder and combines them into a single system prompt. Builds a Gemini Live session configuration (voice, model, and tool schema), signs the session with an HMAC, and requests a single-use ephemeral token from the Gemini API. Returns the token, model, WebSocket URL, signed session metadata, and tool-callback URL to the client via the webhook response. Receives tool-call webhook requests from the Gemini Live client, validates the HMAC signature and expiry, and enforces a per-session tool-call limit. For authorized tool calls, reads order data from Google Sheets to return order status, verifies and appends return/exchange requests to Google Sheets and emails the owner, logs complaints to Google Sheets and emails the owner, or sends owner notifications for unanswered questions. Responds to the tool-call request with a JSON result (or returns an error for missing info, rate limits, unknown tools, or unauthorized calls). Setup Add an HTTP Header Auth credential containing your Gemini API key with header name x-goog-api-key. Connect Google Drive OAuth2, Google Sheets OAuth2, and Gmail OAuth2 credentials used by the workflow. Update the Business Settings values, including owner email, Google Drive folder ID, Google Sheet ID, public tool URL, and a long random HMAC secret. Create or update the Google Sheet with tabs and columns for Siparisler, Iadeler, and Sikayetler matching the fields the workflow reads/writes. Copy the production webhook URLs for /webhook/gemini-live-token and /webhook/gemini-live-tool into your client application and restrict allowedOrigins on both webhooks to your site domain. An n8n automation workflow template by Onur Olgun.

By Pushbuild

Quick overview This workflow receives new lead submissions via a webhook, normalizes the data, scores the lead with OpenAI, sends an acknowledgment email, upserts the contact into HubSpot, and posts a Slack alert when the score meets a configured “hot lead” threshold. How it works Receives a POST request on a webhook endpoint whenever a new lead is submitted. Normalizes incoming form fields into a consistent lead object (name, email, phone, company, message, source) and prepares an OpenAI Chat Completions request. Sends the lead details to the OpenAI API to generate a 1–10 score plus a short reason and summary in JSON. Parses the OpenAI response and attaches the score, reason, and summary to the lead record. Sends a thank-you email to the lead using your SMTP mail server with a booking link for follow-up. Upserts the lead as a contact in HubSpot using the email address as the unique identifier. Posts a message to Slack via an incoming webhook when the lead score is greater than or equal to the configured hot-lead score. Setup Configure the webhook URL in your form tool to POST lead submissions to the workflow’s “New lead” production webhook endpoint. Add an HTTP Header Auth credential for OpenAI (Authorization: Bearer ) and attach it to the OpenAI request. Add an HTTP Header Auth credential for HubSpot (Authorization: Bearer ) and attach it to the HubSpot upsert request. Add an SMTP credential for sending email and set the from address, business name, booking link, Slack incoming webhook URL, and hot-lead score in the Config values. An n8n automation workflow template by Pushbuild.