# Show HN: Friday – a local-first desktop AI assistant that can use tools > Source: > Published: 2026-07-21 10:34:06+00:00 **Friday** v1.0/macOS / Windows / Linux/Private beta # Friday. Your desktop AI copilot for everyday tasks. Chat in plain language and Friday can read and write files, search the web, generate images and video, transcribe your voice, speak back, run on a schedule, and connect to the tools you already use. **Desktop app** Runs natively on macOS, Windows, and Linux. **Your AI providers** Connect your own provider keys and choose the models. **Local by default** Settings, credentials, history, and memory stay on your device. **You stay in control** Friday asks for approval before sensitive actions. **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** **Friday** ## A capable assistant, running on your desktop. ### Chat that acts Ask in plain language and Friday uses tools to finish the job — reads files, drafts documents, searches the web, runs commands, and generates images. ### Local and private Friday is a desktop app. Your keys, settings, credentials, history, and files stay on your machine, not inside a website chatbot. ### Approval before action Writing files, sending messages, and reaching outside services each need your explicit approval before Friday proceeds. ### Speak and listen Dictate instead of typing, transcribe recordings, and have Friday read replies aloud in the voices you choose. ### Skills and MCP Teach Friday reusable workflows and expose new tools through the Model Context Protocol. ### Schedules and checks Run recurring jobs, daily summaries, and health checks on a schedule, even when you are not actively chatting. ## Ask naturally. Approve real actions. STEP 01 Ask### Make a request Type or dictate what you want. Friday understands the goal, constraints, and expected output, then asks only when something critical is ambiguous. STEP 02 Plan### Pick tools It chooses the right files, commands, web search, browser, image generation, skill, MCP tool, subagent, or scheduled workflow for the task. STEP 03 Act### Approve and verify Sensitive actions pause for confirmation. Friday then checks the result and returns an answer, file, image, scheduled job, or other ready-to-use artifact. Unlike a website chatbot, Friday runs on your machine and works through the providers, credentials, and permissions you choose. Friday product brief## Extend Friday with MCP servers. ### Remote MCP servers Connect hosted tool servers over HTTP with optional token or key authentication, then make their tools available to the assistant. ### Local MCP servers Run local stdio tool servers on your machine so domain-specific capabilities plug into Friday without changing the app. ### Any MCP client tool Expose new capabilities to the assistant through the open Model Context Protocol, the single supported way to extend Friday's tool surface. ## Chat with Friday from anywhere. Connect a Telegram bot and chat with Friday from your phone or desktop, with access controls and replies delivered to the same conversation. ### Telegram Bundled bot channel for phone or desktop chat, with allowlists, DM policy, targets, and replies through the same chat. - Allowlist access - Direct-message policy - Same-thread replies ### You decide who can reach Friday. Each channel keeps its own credentials, enabled state, and access rules on your desktop. ## Generation and automation, split by workflow. Assistant The core runtime that plans, asks for approval, uses tools, and returns concrete results. ``` LiveAssistant chatRun agent turns from a clean Markdown chat, with visible tool activity, session history, and permission prompts for sensitive actions. ``` Generation Separate media operators for voice, images, video, and audio so each capability can use its own provider and model. ``` LiveVoice generationDictate with speech-to-text and generate read-aloud replies through the voice provider and model you choose.LiveImage generationCreate images from text descriptions in chat or the image workspace using the provider and model you select.LiveVideo generationGenerate video from text with the provider and model you choose. Results are playable and saved to your media library.LiveAudio generationGenerate music and sound effects from text through the audio provider and model you select. ``` Automation Background tools that run scheduled work and periodic checks without mixing them into media generation. ``` LiveCron schedulerCreate, update, pause, resume, list, and run recurring assistant jobs with their own instructions, provider, and model.LiveHealth checksRun periodic checklist-driven assistant checks and report back when something in your setup needs attention. ``` [View all workflows](/operators) ## Bring your own AI. Mix models by capability. Assistant Speech-to-Text Text-to-Speech Text-to-Image Text-to-Video Music / Audio [View all providers](/providers) Documentation ## Docs that match the product. **8** docs pages **6** sections **Static** Astro content [Product](/docs/overview/) ### What Friday Is Friday is a native desktop AI copilot that chats, acts with tools, and keeps control local. Read docs[Product](/docs/features/) ### Feature Overview The main Friday capabilities: chat, tools, voice, image, video, and audio generation, scheduling, health checks, memory, and personalization. Read docs[Providers](/docs/providers/) ### Providers Friday is bring-your-own AI: add your own API keys and choose providers by capability. Read docs[Extension](/docs/skills-and-mcp/) ### Skills And MCP Extend Friday with reusable skills and Model Context Protocol servers. Read docs[Channels](/docs/channels/) ### Channels Talk to Friday from Telegram and Discord today, with additional channel adapters coming soon. Read docs[Privacy](/docs/privacy/) ### Privacy And Security Friday is local-first, bring-your-own-key, and permissioned before sensitive actions. Read docs[Browse all documentation](/docs) ## Common questions. ## What is Friday? Friday is a native desktop AI copilot for everyday tasks. You chat with it in plain language and it can use tools to read and write files, search the web, generate images, transcribe voice, speak back, run scheduled work, and connect to external services. ## Does Friday send my data anywhere? Friday is local-first. Data leaves your machine only when a task requires an AI provider, channel, MCP server, web request, or hosted service you configured. Provider keys, credentials, settings, memory, and conversation history live locally. ## Do I have to use one specific AI provider? No. Friday is bring-your-own-AI. Add your own provider keys and choose the provider and model for the assistant, speech-to-text, text-to-speech, image generation, video generation, and audio generation. OpenAI-compatible endpoints can be configured too. ## How does permissioning work? Actions that write, delete, publish, send messages, access private data, or touch accounts require explicit approval. Friday shows a confirmation prompt before sensitive work proceeds. ## Can Friday work from outside the app? Yes. Connect a Telegram or Discord bot to talk to Friday from your phone or desktop, with per-channel access control, separate channel secrets, and replies delivered to the same conversation. ## What can be automated? Use the scheduler for recurring assistant jobs such as daily summaries, routine checks, cleanup tasks, and reminders. Health checks can run periodically against a checklist and report only when attention is needed. ## Which desktop platforms are supported? Friday is built with Electron for Windows, macOS on Intel and Apple Silicon, and Linux. The interface supports English and Italian, with light, dark, and system themes. ## Is Friday certified for regulated data? Friday does not currently claim a formal regulated-data certification. Review your own AI providers' and connected services' terms for how they handle data you send them.