Every Featureat a Glance
Everything Plexon can do, from a chat that changes with the job to 192 specialists that work side by side on it.
How It Thinks
AI Chat & Modes
Five specialized modes for every workflow: Code, Plan, Ask, Auto, and Orchestrator.
Plan Mode & Saved Plans
The AI explores read-only and writes one implementation plan you approve. Every plan stays browsable, ready to implement or edit.
Multi-Agent Orchestrator
Break complex tasks into parallel subtasks delegated to specialized AI agents.
Team Coordination
Persistent multi-agent teams with mailboxes, shared memory, and member wake-up across sessions.
192 AI Agents
Specialists for every kind of work, each with its own expertise and its own way of going about it. Describe the one you want in chat and Plexon writes it for you.
Personas
Twelve roles to start from, each arriving with the right specialists, skills, plugins and tools already switched on. Make one of your own and the AI drafts the description and the opening suggestions for you.
Skills & Tools
Teach Plexon something once and it keeps it: how you like a document laid out, the steps you follow, the tools it should reach for. Describe a new one in chat and it writes it for you.
Self-Improving Skills
Plexon learns from each session and quietly improves its skill library. No manual editing. A curator agent reviews what happened, saves new skills, refines existing ones, retires the stale, and protects what you've pinned.
Quality Harness
A quality layer that makes the model verify its own work. It checks facts before using them, writes a spec before multi-step tasks, hunts the eleven shortcuts that fake completion, and files a pass or fail verdict after every change.
Prompt Suggestions
AI predicts your next message after each response. Press Tab to accept or just start typing.
Side Questions
Type @btw and ask about the conversation while the AI keeps working. The answer streams into a side card and never enters the session history.
Session Branching
Rewinding a chat deletes everything after the point you rewind to, so a long piece of work is the price of one experiment. Branch at any message instead and the conversation up to that point opens as a second chat, with the first one untouched. A line marks where the inherited part ends and clicks back to where it came from, each branch is named after the question at its branch point rather than the answer, and the files the original produced are shared rather than copied.
Adaptive Routing
Plexon notices which suggestions you take and which agents finish the job, then puts the next ones in that order. Until it has watched enough to be right it does nothing, so the order on day one is exactly the order you get today. What it learned is one file on your machine, shown with the run count behind every number, and one button puts it back.
Knowledge Graph
Plexon connects the notes it keeps for you, so a question about a client is one question instead of six file opens. Every fact answers with the note and the line it was read from, and whether the note stated it or Plexon worked it out. Two spellings of one name are proposed for merging and never merged on their own, a contradiction stays on screen rather than being settled for you, and the whole thing is derived from your markdown, stays on your machine, and clears in one action.
Decision Tools
Routes, container loads, shift rosters, and budget splits are searches, and an answer written from intuition looks exactly like a correct one. Plexon computes them instead, and reports the seed that reproduces the result. Forecasts come back as a range with the odds attached rather than a single number.
Projects
Keep the work you are actually doing, in one place per machine rather than per repository. A project holds tasks; a task points at any folder, nests as subtasks five levels deep, repeats on a schedule, names who owes it, and links out to a Jira or GitHub issue that opens in your browser. A board with a column per status and cards you drag across it. Every record is a markdown file you can edit by hand, and they are drawn on the memory graph beside the people and notes they relate to.
Checklist
On a multi-step request the AI keeps a live checklist per chat, with a progress line above the message box showing what it is doing right now. Between turns the list is yours: check steps off, add, delete, or reorder them, and the AI honors your changes on its next turn. Send a step to Projects to keep it beyond the chat. Finished turns close their last dangling item; Stop never leaves a fake spinner.
Automation
Dynamic Workflows
Lay out the steps one by one, or describe what should happen in chat. Steps can call one of your apps without spending anything on AI, pull details out of a message, run an agent or a role, branch on a condition, wait for a reply, or stop and ask you. It runs on your computer, keeps a record of every run, and never sends the same thing twice. Ships a Greet new contacts on WhatsApp example.
Richer Hook Events
Six lifecycle events instead of two: pre/post tool use, tool failure, user-prompt expansion, config change, and pre-compact. A new agent hook type dispatches the payload to a sub-agent, and personas can now bundle their own hooks that activate with the persona and read the finished session transcript. No code needed: Create with AI writes the script for you, and starter templates cover every event.
AutoDream Memory Consolidation
Your AI learns across sessions and keeps its knowledge organized automatically. A background agent deduplicates entries, removes stale facts, resolves contradictions, and maintains a clean MEMORY.md index. The same run turns your thumbs feedback into rules it follows afterwards.
Flows
When Plexon repeats an action across many things, or checks something on a timer, it puts a card in the chat. The card counts what went through and what did not, with the reason for each, and gives that job its own pause and stop. Hold one job without ending the conversation. Run several at once, skip a wait, or retry only what failed.
Connections
App Connectors
67 branded app integrations: Slack, Notion, GitHub, WhatsApp, Telegram, Twilio, n8n, Zapier, Pipedream, Make, Vercel, Netlify, Cloudflare, Supabase, Stripe, Salesforce, Zendesk, Intercom, Keragon, FitMetrics, Airtable, Brave Search, Perplexity, and more. Turn one on with your own account and the AI works inside it.
API Connections
Turn any HTTP API into AI-callable tools. Import an OpenAPI 3.x spec or a Postman collection, or build endpoints by hand; pick API key, Bearer, Basic, or OAuth 2.0. Attach reusable middlewares to inject identity, add headers, or reshape responses. Bundle a connection into a persona or share it through the marketplace. SSRF defense with localhost allowed by default, per-host allowlist, and a per-endpoint test runner.
Plexon Provider
One subscription, every capability. Automatic model routing, key management, and multi-task agents.
Fireworks AI
Use open models through Fireworks from inside Plexon: GLM 5.2, Kimi K2.7 Code, Qwen 3.7 Plus and more. Room for very long conversations, a faster serving option, and the longer a chat runs the cheaper most of it bills. Use your own Fireworks account.
Custom Providers
Point Plexon at any OpenAI- or Anthropic-compatible endpoint: local LLMs via Ollama, LM Studio, and vLLM, Azure OpenAI, and alternative backends like Moonshot’s /anthropic endpoint. Per-model protocol and capability flags, Test Connection, Fetch Models, JSON import/export. Premium plan.
Data Sources
Load a local spreadsheet or CSV once and use its columns everywhere. Workflows loop the rows, dashboards chart them, and a built-in tool reads them from any prompt. Ask the assistant for a new list, or to add, fix, or drop rows in one you have, and the panel updates as it goes. Keep a source as a fixed snapshot or a live read of the file, preview every row in a paged and filterable grid, and move a source between this project and all projects.
Unified Marketplace
One Browse window for everything you can install: skills, agents, plugins, personas and tools. Drop in a zip, paste a link, or click any item in a marketplace you have added. Plexon works out what it is and installs it the right way. A single bundle can mix all of them, private sources stay encrypted, and nothing locks you in.
Plugins
Fourteen bundled plugins: dev workflows, office authoring, Google Workspace, Microsoft 365, GitHub, graph-aware code review, databases, browser testing, design-language extraction, GEO-SEO, marketing, nutrition (dietitian), and stock-market research. Install more from GitHub.
What You Get Back
Dashboards
Boards of live widgets fed by your tools, the AI, a workflow, or static text. Stat tiles, charts, gauges, tables, and HTML, refreshed on a timer or in the background. Add a parameter like a city and the whole board follows. Draft one in chat or build it by hand. Ships a Weather & City board and a Global Intelligence board built on free public data.
Dynamic UI Widgets
Answers come back as things you can use: cards, grids, tables, timelines, checklists you can tick, and forms you can fill in, instead of a wall of text. Thirteen kinds in all.
Office Integration
Read, write and edit your Word and Excel files straight from the chat. Nothing to set up, works on every platform, and edits show up as tracked changes while you have the document open in Word.
File Organization
Ask Plexon to tidy a folder and it does the moving itself: files renamed, sorted into folders, copied where you want them, and the ones you are done with cleared out. Anything it removes goes to the recycle bin, so nothing is gone for good. It will also tell you what is taking up the space before you decide.
Screenshot Cleanup
Ask Plexon to clear the screenshots out of a photo folder or a phone backup. It reads each file to work out what it is, so it still finds them after an export has renamed everything to IMG_9084.PNG. Camera photos are recognised separately and left where they are, and what goes, goes to the recycle bin.
Document Reader
Point Plexon at a Word doc, Excel or CSV file, PowerPoint deck, PDF, EPUB, HTML page, or Outlook .msg email and it reads the contents as Markdown you can ask about. Ask about a report, a budget, or a slide deck without converting anything first. The reader installs itself on first use, with nothing to set up.
Visual & Voice
Talk to it and it talks back, hands free, taking turns like a phone call. It listens on your own machine, answers in one of 400+ voices across 80+ languages, lets you cut in mid-sentence, and keeps a written record of what was said. You can also paste in a picture and ask about it.
Image Generation
Generate images inline with Nano Banana 2 Pro (Gemini). Describe what you want, get it instantly.
Video Generation
Create videos from text or images. Product demos, animations, and visual content generated inline.
Music Generation
Describe a song and Plexon writes the words, sings them, and hands you the file. Instrumentals and covers too.
Video Composition
Turn your photos, recordings, and generated clips into one finished, narrated video with brand cards and timed captions. Renders on your computer. No daily limit.
Telegram Bot
Control your AI remotely via Telegram. Chat, manage sessions, send images, and monitor tools.
Slack Bot
Talk to Plexon from Slack. Message it directly or @mention it in a channel, send it images and files, and have your routines post results back. Works on Apple Watch through the Slack app.
Session Sharing
Save any conversation as one encrypted file you can send someone. They drag it into their own Plexon, or open it from a link, and it arrives looking exactly as it did for you. They can carry on from where you left off. No cloud, no accounts.
Live Share (Premium)
Share a running session and work with the AI together in real time. You send an invite code, they send one back, and the two apps connect straight to each other: nothing reaches a Plexon server, not even the connection details. Guests wait for you to admit them, and you both see four safety words to check first. They start read-only; promote someone to let them drive. Turn-taking stops prompts from colliding, and tool output that could expose the host machine stays hidden from guests. Both apps have to reach each other directly, so the same network is the reliable case. Premium plan only.
Session Lock & App Lock (Pro)
Lock a chat with a passcode. Its messages are encrypted with a key derived from that passcode, so nothing reads them without it. Unlock with Touch ID on macOS or Windows Hello on Windows. Locked chats still show in the sidebar by name with a lock badge, and stay out of exports and live shares until you open them. A separate app lock asks for a passcode before the Plexon window becomes usable on launch, and an optional limit on wrong PINs erases everything Plexon keeps on the machine once it is reached. Both locks are local: no passcode, salt, or key reaches a server, and a forgotten session passcode cannot be reset. Turning a lock on is Premium; unlocking, changing a passcode, and removing a lock work on any plan.
Move to Another Machine
Move your whole Plexon to a new computer. One encrypted file carries your personas, skills, agents, connectors and the credentials they use, plugins, workflows, dashboards, memory, journal, cloned voice, and chats. Install Plexon on the new computer and click "Bring my setup from another machine" on the sign-in screen. You pick the passphrase that seals the file and Plexon stores it nowhere, so it offers a recovery code beside it. No cloud and no account sync. Import replaces what is on the machine after naming the exact paths it removes, and keeps the old setup as a restore point for a week. Works between Windows, macOS, and Linux, and the same file doubles as a snapshot you can restore later.
Cloud Backup (Google Drive)
Connect your own Google Drive and Plexon keeps your setup and every project's chats backed up to a hidden folder there: settings, skills, agents, personas, connectors with the credentials they use, workflows, dashboards, memory, journal, notes, and your cloned voice. Everything is encrypted before it uploads, and only what changed uploads. A fresh install restores it all from "Restore from Google Drive" on the sign-in screen, and a project folder whose chats are in the backup gets them back when you open it. Your Google sign-in is enough to restore by default; an optional passphrase makes Google hold ciphertext only. The backup counts against your own Drive storage, and revoking Plexon's access in your Google account deletes it.
Cross-Chat Messaging
The chats you have open on one project keep each other informed. A change in one reaches the others, they share a board of decisions and claims every chat reads, and @chat: names a specific one. Capped at two hops and twenty messages per pair so two chats cannot talk forever.
Pinned Messages
Pin any message from either side of the chat. The Pinned view in the sessions sidebar searches across every project on the machine and drops you back on the exact message.
Command Palette
One search box in the title bar, or Ctrl+K, over everything Plexon has: skills, agents, connectors, personas, workflows, dashboards, chats, plans, memory notes, the people and facts in your knowledge map, pinned messages, all of the settings, and every panel. Results rank as you type and the matched letters are picked out. Type a product name to install a connector, or describe a setting in your own words and land on the exact control. It acts as well as finds: switch mode or theme, register a marketplace source, open what is new. Type / to search the words in the chat you have open, and when nothing matches it offers to send what you typed as a message. Ranking runs inside the window, so nothing waits on a server and no file contents are ever read.
Customization
Ask the assistant to explain or change any setting. 13 themes, 8+ providers, voice I/O, Desktop Automation, and granular controls.
Migrate from Claude Code
Bring your Claude Code setup across: agents, connected tools, automations, plans, memory, and every conversation. Plans land in the project each one was written for and keep their approved or draft status, and everything imported is labelled with where it came from.
Developer
Publish Sites
Put the website in your project on a real address. Preview it exactly as a static host will serve it, so a broken base path or a dead deep link shows up before anyone else sees it. Share a temporary link with no account while you are still working. Then publish to GitHub Pages: Plexon creates the repository, writes the deploy file for you to approve, pushes, follows the build, and fetches the published address itself before it says the site is live. It checks for the failures that leave every status green, adds your own domain with the exact records for the kind of name it is, and can run the whole thing from a workflow on a schedule.
Code Intelligence
In a Go, Python, TypeScript or Rust project, Plexon runs the same language server your editor would and asks it about the code: where something is defined, everywhere it is used, what implements or calls it, what a file declares, and what is currently broken. Ask by name, so no line and column numbers change hands. When it edits source, the errors it just introduced come back with the edit and get fixed in the same reply. Servers install in the background and nothing needs configuring.
Developer Tools
Source control, automation hooks, and prompt engineering built into the workflow.
Parallel Agents Workspace
Fan out one prompt across N agents in isolated git worktrees. Live tile grid with ff / squash / cherry-pick merge per tile, whole-run cancel that keeps partials, and pre-flight safety checks that never let a failed merge destroy your scratch state.
Diff Review Queue
Cross-session aggregator for every parallel-agents tile waiting on review. Merge with ff / squash / cherry-pick or discard from one panel; toast + WebAudio chime fires when a remote tile finishes.
Spec-Driven Mode
Living markdown contracts (AGENTS.md + .specify/) drive constitution → specify → plan → tasks → implement, with a verifier subagent at every step. Title-bar status pill is the entire UI: one click stages a grounded prompt and auto-sends. GitHub Spec Kit-compatible, same files work in Cursor, Copilot, Claude Code, and 30+ AGENTS.md-aware tools.
Graph-Aware Code Review
Persistent Tree-sitter knowledge graph of your repo. Reviews read the slice a diff touches (not whole files) for roughly 8× fewer review tokens.
Embedded Browser
A split-pane webview docked next to the chat. Point at any DOM element on your local dev server (localhost:3000, Vite, Next.js, staging) and the AI gets the exact selector, viewport, outer HTML, computed styles, and a cropped screenshot inline in the prompt. Four-tier selector strategy, zero browser extension, zero setup.
GitHub Auto-Review
Point Plexon at a repository you cloned and it watches GitHub while you work: every new issue gets a drafted reply, every new pull request gets a review with a verdict, and both wait in the Approvals panel until you press Post. Set a start date so old threads stay untouched. Turn on Always for one kind of action on one repository and Plexon handles it alone, with a daily cap and one click to withdraw it. It can also write the fix: a patch in a throwaway copy of your repository, your own tests run against it, and the diff shown to you before anything is pushed.
Shell Commands
Start a message with ! to run it as a shell command in the session's working directory. The output renders as a terminal card in the chat, joins the session history, and rides into the AI's context on your next message. Same per-OS shell routing as the AI's own command tool, and a run past 90 seconds moves to the background instead of being killed.
Design Extract
Extract the full design language from any live URL (W3C DTCG tokens, typography, spacing, shadows, components, WCAG scoring) and get 8 ready-to-use handoff artifacts for Tailwind, shadcn/ui, Figma, React, and vanilla CSS. Bundled into the Designer persona.
Marketplace Developer Mode
Build your own marketplace from inside Plexon and collaborate on it without git. One-click Sync publishes your edits and pulls your teammates' changes, conflicts resolve per file (keep mine, take theirs, or AI merge), workflows and dashboards write back into the folder on save, and every install updates itself on launch. Scaffolder, live-reload editing, Inspector validation, version bump, per-marketplace Source Control. Plaintext on GitHub.
See it on your own machine
Every feature on this page ships in one download. Pick a persona and try it on real work.
