Morphy, your assistant
The chat that's always on your canvas. It builds cards, opens what you need, and starts and keeps track of your other agents.
Morphy is the chat that's always on your canvas. It's a full agent, like any other chat, but it also knows Morphy itself: it can build cards, open files and websites, start and manage other chats, and bring the right ones together when you want to work on something.
It can't be closed or deleted, so it's always one click away.
What to ask it
| Ask for | Morphy will |
|---|---|
| A dashboard, tracker, overview or tool | Build a card on your canvas |
| A question or a task | Answer it, or just do it |
| "Open the Stripe dashboard" / "open my notes file" | Give you a link that opens it in your browser or its own app |
| "Where's my invoice tracker?" / "Which chat reviewed the API?" | Link to that card, chat or project: click it to go there, even from focus mode |
| "Open three Codex chats in my website folder and have them review the checkout" | Start the chats and send them the prompt |
| "Which chats are done?" | Check every chat and tell you what's running and what finished |
| "I want to work on the website" | Open the chats and cards for that topic in focus mode |
| "Make a School project and put my resit notes in it" / "Line my projects up in a grid" | Create, rename or remove projects, move chats and cards between them, and rearrange their groups on the canvas |
A Claude doc, or a claude.ai/artifact/… link | Make it, or read it, with Claude Docs (see below) |
| "Find the contract in my Drive" | Look in Google Drive, if you connected it on claude.ai (see below) |
| "Make a card of my open pull requests" / "Turn my Notion roadmap page into Linear issues" | Use a tool you connected, like GitHub, Notion or Linear (see Connect your tools) |
Claude Docs and your claude.ai connectors
When Morphy runs on Claude, signed in with a Claude account that has the Claude Docs connector, Morphy can use it: ask it for a Claude doc, or paste a claude.ai/artifact/… link and it reads it. The doc lives on claude.ai and opens in your browser. Anything else Morphy makes for you, like a report, a list or notes, still goes on your canvas as a card or note.
Morphy also uses the other connectors you connected on claude.ai, like Google Drive or Google Calendar. Ask it to "find the contract in my Drive". It uses only the ones that are connected; one you added on claude.ai but haven't signed in to yet stays out until you do. What a connector can do is up to claude.ai.
Morphy finds your connectors by itself shortly after it starts, and checks again about once an hour, so one you connect on claude.ai shows up within the hour. A message you send before then runs without it; the next one has it.
A Claude chat working in a project folder has all of your claude.ai connectors, plus the skills you saved on claude.ai, like Claude Code in a terminal.
Connect your tools
Settings → Morphy Agent → Connections lists the services Morphy can use as tools: MCP servers, the standard way AIs reach your GitHub, your docs, your issues or your payments. Each has a Connect button:
| What Morphy gets | How it connects | |
|---|---|---|
| GitHub | Your repositories, issues and pull requests | A fine-grained personal access token |
| Notion | Your pages, docs and databases | Sign in through your browser |
| Linear | Your issues, projects and cycles | Sign in through your browser |
| Sentry | Errors and issues in your apps | Sign in through your browser |
| Context7 | Up-to-date docs and code examples for any library | Nothing (a free key raises its rate limit) |
| Supabase | Your databases, tables, logs and edge functions | Sign in through your browser |
| Cloudflare | Your Workers, DNS, storage and the rest of your account | Sign in through your browser |
| Stripe | Payments, customers, subscriptions and invoices | Sign in through your browser, or a restricted key (rk_…) |
| Apollo.io | Prospects, companies, contacts and sequences | Sign in through your browser |
| vidIQ | YouTube research: channels, videos, trends and keywords | Sign in through your browser |
| Higgsfield | Images and videos from its AI models | Sign in through your browser |
| fal | Image, video and audio models to make media with | Sign in through your browser |
| ElevenLabs | Your voice agents, and speech from text | Sign in through your browser |
| Dropbox | Your files and folders, and what you share | Sign in through your browser |
| Metricool | Social media analytics, posts and scheduling | Sign in through your browser |
| Wix | Your sites and what runs on them: store, bookings, blog, contacts | Sign in through your browser |
| Figma | Your designs, while the Figma desktop app is open | Nothing (it needs a Dev or Full seat on a paid Figma plan) |
| Obsidian | Your notes and your vault, while Obsidian is open | The API key of its Local REST API with MCP plugin |
What an agent makes with Higgsfield, fal or ElevenLabs, and what it looks up with vidIQ, uses your credits there, as in their own apps. Metricool works on any of its plans, Free included, within that plan's limits. A Supabase, Cloudflare, Dropbox, Metricool or Wix sign-in lets agents change things in your account, not only read them (move or share files, schedule posts, publish your site, edit products), so keep production out of their reach or keep those chats in Ask first. ElevenLabs can also make, change, test and delete your voice agents: a change goes live at once and a deleted agent is gone, so it starts in Morphy only and is best kept in Ask first. Its speech comes back as a download link that expires, so ask for the audio to be saved in your project if you want to keep it. ElevenLabs connects to its global address; if your workspace is in the EU, India or Singapore, add it as an MCP server of your own (below) with your region's address instead: https://api.eu.residency.elevenlabs.io/v1/mcp, https://api.in.residency.elevenlabs.io/v1/mcp or https://api.sg.residency.elevenlabs.io/v1/mcp. Shopify isn't in the list: its server lets in only the AI apps Shopify has approved, and Morphy isn't one of them yet.
Connect works right in the list, with no page in between. For a service you sign in to, your browser opens the service's own page; sign in there if you aren't already, click Allow, and the row says Connected. That's it: Morphy keeps the sign-in on your computer, sealed, and renews it by itself as it runs out, so you don't sign in again unless the service ends it (you remove Morphy's access there, for instance). Then the row offers Sign in again. While Morphy waits for you in the browser, the row says Signing in…, with Cancel. If a service lets in only the apps it has approved and Morphy isn't one of them, the row says so right away, before any page opens. Context7 and Figma need nothing, so Connect adds them at once. GitHub asks for its token in the row: paste it and click Connect (Make one opens the page where you make it).
Obsidian runs on your computer, not online, so agents reach it through an Obsidian plugin. In Obsidian, open Settings → Community plugins, install and enable Local REST API with MCP, then in its settings turn on Enable non-encrypted (HTTP) server and copy the API key. In Morphy, click Connect on Obsidian's row, paste the key and click Connect (Get the plugin opens the plugin's page). Obsidian has to be open for agents to reach your notes. While it's closed, the row says Can't connect and chats go on without it. Once you open Obsidian, a Claude chat has it again from your next message. A chat on Codex, Grok, Copilot or OpenCode that started while Obsidian was closed gets it the next time Morphy starts. The plugin's server listens only on your own computer and needs the key for every request. Agents can write, move and delete notes as well as read them; a chat in Ask first asks before each change.
Figma also runs on your computer: agents read your designs through the Figma desktop app's own MCP server, which comes with a Dev or Full seat on a paid Figma plan (Professional or above). In the Figma desktop app, open a design file, switch to Dev Mode (Shift+D) and, with nothing selected, click Enable desktop MCP server in the right sidebar. Then click Connect on Figma's row in Morphy. It starts in Morphy only; to build your designs in code, open its page and give it to the chats that do that under Use in. Select a frame in Figma and ask a chat to build it, or to use your file's variables and styles. This server reads your designs; it can't draw on your canvas. Figma has to be open: while it's closed, the row says Can't connect and chats go on without it, as with Obsidian. Figma counts each read against your seat's limit (200 a day on a Professional plan). Figma's online server, which can also draw on your canvas, lets in only the apps Figma has listed, and Morphy isn't one of them yet.
A connected one opens its own page: click its row. It says how the connection is doing (Connected · 41 tools, a key the server refused, or an address that didn't answer), with Check again, which chats get it (Use in), Turned on, its key (Change), its tools and Remove. Sign out there forgets a sign-in and tells the service. If the browser tab got closed or lost while signing in, Open the page again brings it back. On Stripe's page, Use a restricted key instead swaps the sign-in for a key, and Sign in through your browser instead swaps it back.
A service that isn't in the list: the last row, Missing a service?, asks us to add it. Click Request, type its name and click Send (or press Enter). It reaches the Morphy team the way feedback does, with your name and email; the row says Sent once it's on its way.
An MCP server of your own: click Add at the top right of Connections (or press Ctrl+K, ⌘K on a Mac, and pick Add your own MCP server…). Paste its web address (and a key, if it needs one), or the command that starts it, like npx -y @modelcontextprotocol/server-memory. A server that signs in through the browser (Atlassian's, Canva's and many more do) opens its sign-in as soon as you click Add. Add MCP servers from your other apps… (also in Ctrl+K) lists the servers you already set up in Claude Code, Codex, Claude Desktop, Cursor, Gemini CLI, Windsurf or VS Code: add one and Morphy copies it, keys included; one you signed in to there opens its sign-in for Morphy, which signs in on its own. Those apps aren't changed. Servers you add this way are listed under Morphy's own.
Which chats get it. A connection always goes to Morphy's own chat. Use in, on its page, says which other chats get it too:
- Morphy only: no other chat.
- Projects: the chats in the projects you switch on. Chats in Home don't get it. Use this for a connection only some of your work needs, like a whiteboard for two projects, and keep one that can change things, like Cloudflare, in Morphy only.
- Every chat: all of them. Context7 starts here, since coding chats are where it helps.
It works whichever AI a chat runs on: Claude, Codex, Grok, Copilot or OpenCode (Antigravity can't use MCP servers yet). Renaming a project keeps its connections; deleting one takes it off them. A chat you move to another project gets that project's connections. When you add a server of your own, Use in every chat gives it to every chat at once; pick projects on its page afterwards. A chat that already has the same server from your own setup keeps just its own. A change reaches each chat with its next message, so nothing stops mid-answer. Turned on off keeps a connection in the list without giving it to any chat.
A connection's tools follow the chat's permission mode: in Ask first, the chat asks before every call. Keys go into Morphy's .env file (Keys → Morphy's .env file, under Connections), never into the list itself, and Remove takes a connection's key out of it too (or signs out of it). Give a connection a key with no more access than your agents need, like a read-only GitHub token or a restricted Stripe key.
A signed-in connection's sign-in never goes to an AI: your chats reach the service through Morphy itself, which adds the sign-in to each request, and only a chat you gave the connection to gets through. One sign-in serves every chat on every AI.
Some services sign in through the browser but don't let apps register by themselves (GitHub is one); Morphy says so when you check one, and if the service gives you a key you can add it with the key instead. A chat shown as a terminal, or working in a folder on another computer, doesn't get your connections.
Run other agents
Morphy can start chats with any connected AI, in any folder, and send them prompts. Ask it in plain words:
Spawn two Claude chats and two Codex chats in
C:\code\shop, and ask each one to find the slowest page.
Send the Codex chat that's reviewing the API: "also check the error handling".
Compare what the four chats came back with and tell me which approach to take.
The new chats appear on the canvas to the right of what's already there, and the prompts Morphy sends show up in them as normal messages. You can take over any chat at any time: just type in it.
When you ask what's going on, Morphy checks every chat fresh (it never answers from memory) and reports per chat: which folder, whether it's running or done, and how long ago. Ask it to compare what the chats came back with and it tells you how their answers differ.
Morphy only sends a message to a chat that isn't busy, and it won't close chats it didn't start without asking you.
Ask for a team and Morphy builds it as a flow (Pro):
Get a team of four on which cache the API should use, and have them argue it out.
Morphy first asks what matters (the goal and the limits). Then it makes a flow with that brief, for example a council: four chats answer from different roles, on different AIs where you have them, and a fifth writes the combined answer into a note. Morphy starts it and tells you the result in a few lines. Ask it for a team, a review, a pipeline or something that runs every morning the same way.
Focus on a topic
Tell Morphy "I want to work on X" and it remembers which chats and cards belong to that topic. Next time you say it, Morphy checks they still exist and opens them together in focus mode, with the most important one largest. Ask for a view if you like: "in stage" puts the main chat big with the others watched beside it, "in a strip" lines many chats up in columns you scroll through. Morphy remembers the view you asked for with the topic.
Choose the AI behind Morphy
Morphy runs on Claude by default. Switch it in Settings → General → Morphy runs on: any connected AI can run it. You can't switch while Morphy is working; stop it first.
If Claude isn't connected and you've never chosen, Morphy switches itself to Codex, Grok, Copilot, Antigravity, OpenRouter or OpenCode, whichever you have connected (OpenCode last: it only needs to be installed). It never picks Ollama, Hermes Agent or Junie by itself: to run Morphy on one of them, choose it here.
Teach it how you work
Morphy works from three files. Find them in Settings → Morphy Agent → Instructions:
| File | What it's for | |
|---|---|---|
AGENTS.md | About you, and how Morphy works with you | Yours to edit |
CANVAS-USER.md | How you like your canvas and cards (style, habits, cards you keep) | Yours to edit |
CANVAS-APP.md | What Morphy can do in the app | Built in, read-only |
AGENTS.md is the name Codex, Copilot, Gemini and most other AIs look for, so it's the same file whichever AI Morphy runs on. Next to it, Morphy keeps a one-line CLAUDE.md that points Claude at it; leave that one as it is.
Edit opens a file in your Markdown editor (or Notepad on Windows, TextEdit on a Mac). Or skip the editing and just tell Morphy: "remember that I prefer short answers", "make my cards compact from now on". It updates the files itself.
Keep these two files short: on Claude, Morphy reads them before every message (other AIs read them when a conversation starts). Morphy's built-in card style, and its guides for bigger jobs (building cards, running other chats, projects, focus), are part of the app and read only when a task needs them. What you write in CANVAS-USER.md wins over the built-in style. For longer notes, put a separate .md file in Morphy's folder and mention it in AGENTS.md: Morphy opens it when a task needs it. A line with only @ and its name (@PROFILE.md) sends that file along with every message instead, on every AI, so keep those few and short. Agent folder → Open takes you there: add notes, scripts and tools, and Morphy can use them.
See Your files for where everything is stored.
Your Claude Code skills
If you use Claude Code, Morphy knows your skills and plugins there. Each time it starts, it reads their names and descriptions from ~/.claude (never your hooks, your CLAUDE.md or any key). When a task fits a skill, Morphy opens it and follows it, whichever AI it runs on. You can also pick one in the / menu, under Claude Code. When a plugin needs its own tools, MCP servers or hooks, Morphy hands the task to a Claude chat, which loads your whole Claude Code setup, and tells you what came back. See Your Claude Code setup.
Morphy's folder
Morphy works in its own folder, where its instructions and scripts live:
| On Claude | On other AIs | |
|---|---|---|
| Windows | %APPDATA%\Morphy\assistant | %APPDATA%\Morphy |
| Mac | ~/Library/Application Support/Morphy/assistant | ~/Library/Application Support/Morphy |
Leave it there: if you point Morphy at another folder, it loses its built-in instructions and can no longer build cards or manage chats. For work in a project folder, start a separate chat there instead.
