YOUR SUBSCRIPTION. YOUR AGENTS. YOUR TOOLS.

Your AI.
Your tools.
One shared workspace.

Bring your AI subscription. Connect your tools through MCP. Keep your agents, collaborators and progress together in DevGhouse.

Claude, Codex or Kimi. Start with the account you already use.

Connect through MCP↗Agents in one conversation↗Routines on a schedule↗Your computer, paired↗Work local
01 / MODEL CONTEXT PROTOCOL

Your tools, within your agent’s reach.

MCP lets your assistant work with the tools you connect to DevGhouse. Choose the realm, project and permissions. Give it context once, then keep working in the client you prefer.

Claude Code
claude mcp add --transport http --scope user devghouse \
  https://mcp.devghouse.com/mcp \
  --header "Authorization: Bearer YOUR_DEVGHOUSE_TOKEN"

Swap YOUR_DEVGHOUSE_TOKEN for your real token.

Your AI

  • Claude Code
  • Codex
  • Cursor
  • Claude.ai

DevGhouse MCP

One token. Scoped to your workspace.

1,500+ tools

Your connected services

  • GitHub
  • Trello
  • Vercel
  • Claude
  • Kimi
  • OpenAI (GPT)
  • Gemini
  • Qwen
  • Alibaba Cloud
  • DeepSeek
  • ElevenLabs
  • X
  • +27 more

Your rails

CONNECT ONCE, THEN ASK

Your agent, working
in your workspace.

  1. CopyTake the command from the connect page.
  2. AskTell your assistant what you need.
  3. WorkIt calls DevGhouse tools and reports back.

Every call stays inside the workspace your token belongs to, with your permissions.

The DevGhouse connect page copies the Claude Code command, then Claude Code drafts a blog post through DevGhouse MCP tools
The real hub playing scripted example data, with a placeholder token. No post was drafted.
02 / AI AND COMMUNICATION

Your agents work.
Everyone stays in the conversation.

Start with a brief. Keep the work, feedback and next decision in one thread. Bring in a collaborator when you need another pair of eyes.

ONE REQUEST, START TO FINISH

Ask in plain words.
Watch every step.

  1. ScoutReads the project, its pages and its tools.
  2. PlanSays what it will change before it changes it.
  3. BuildWorks in its own branch.
  4. TestOpens the result in a browser.
  5. ReviewHands the change to a second agent.

It comes back with the Reviewer’s verdict and a link to sandbox, your staging site.

A DevGhouse thread where the agent’s steps, the Reviewer’s PASS and the sandbox link follow one request, then the highlight updates
Boris Zarinski Boris+Bak Fun Bak+ Their agents DevGhouse realm
DevGhouse / AgentsIllustrative demo
Illustrative DevGhouse conversation with Boris, Bak and a local agent
A request. A second pair of eyes. One shared result.
THE HOOKS KEEP IT TOGETHER

From your terminal
to the shared thread.

  1. 01
    Keep the original request

    The hook mirrors your words into the right conversation, under your name.

  2. 02
    Bring replies back

    At work boundaries, the local session checks the thread for new messages.

  3. 03
    Make the result easy to find

    Agents post progress and evidence. Highlights gives the team the current result.

Both clips are the real frontend playing example data. In the Agent Mode clip the run, the Reviewer’s verdict and the sandbox link are scripted; the conversation clip uses invented messages. No backend actions are performed.

03 / AUTOMATED SCHEDULING

Set it once.
It runs on time.

Routines give an agent a standing job. They run on their schedule, keep every result, and wait for you to read them.

YOUR AUTOMATIONS

Every routine shows
when it runs next.

  1. ScheduleDaily, weekly or on a webhook or form.
  2. RunEach run gets its own thread and result.
  3. CheckSee the next run, or run it now.

The same agents and tools as your chats, on a timer.

Routines with their next runs; one is opened and run now until its result lands
The real hub playing scripted example data. The schedules, runs and result are examples.
04 / CONNECT YOUR DEVICE

Pair your computer.
See it in Devices.

DevGhouse Desktop pairs with a short code. Your computer then shows in Devices with the AI accounts signed in on it and the MCP servers it can reach.

ONE CODE, ONE COMPUTER

From a code
to a working computer.

  1. PairGet a code here and enter it in DevGhouse Desktop.
  2. SeeDevices lists the computer and its status.
  3. UseIts AI accounts and MCP servers are ready.

Pair a computer

A pairing code is claimed by DevGhouse Desktop, then Devices shows the computer with its AI accounts and MCP servers
The real hub playing scripted example data. The computer, accounts and servers are examples.

Local sessions need the connected computer online. Device setup and engine availability are checked in your workspace.

05 / WORK LOCAL

Your computer does the work.
You steer it from the hub.

Start a chat on a paired computer and Claude or Codex answers there, with that computer's own logins and folders. The conversation stays in app.devghouse.com.

NEW CHAT, ON YOUR COMPUTER

Pick where it runs.
Then just ask.

  1. WherePick a paired computer in New chat.
  2. EngineClaude or Codex, signed in on that computer.
  3. AnswerIt runs there, in your folders, with its own logins.

The computer needs DevGhouse Desktop running. When it is off, pick DevGhouse servers instead.

New chat with Where set to a paired computer, Claude and Claude Opus; the thread runs on that computer and the answer comes from it
The real hub playing scripted example data. The computer, the run and its answer are examples.
06 / WEBSITES

Your site.
How Google shows it.

Every website belongs to a project, in the same workspace as your agents. How it shows in search is set from there.

SEARCH SETTINGS

See how Google
will show your site.

  1. WriteSet your site's search title, description and share image.
  2. PreviewSee how Google will show it while you type.
  3. SaveYour site reads it from DevGhouse through its website link.

Websites

devghouse.com in the workspace: its search title and description being written, with the Google preview below
The real hub replaying recorded devghouse.com settings, mid-edit.
07 / CRYPTO

Launch a coin on pump.fun.
From a chat, a form or Telegram.

An agent launches it over MCP, or you click it through on crypto.devghouse.com. The coin goes out on pump.fun on Solana with its first buy, and the creator fee split is set right after the mint. On the local test chain below, DevGhouse also reads the split back from the chain.

Agent chatClaude, Codex or Cursor over MCP
Launch formcrypto.devghouse.com, with Mayhem and buy-back options
Telegram bot/coin, pay with Solana Pay
DevGhouseChecks the wallet and the chain, then signs the launch
pump.fun on SolanaMint and first buy, then the fee split right after
50 · 50
Creator fee splitUp to 10 wallets, set right after the mint; pump.fun then locks it
PumpSwappump.fun moves the coin when its curve fills; DevGhouse tracks the progress
ASK YOUR AGENT

The launchpad tools
your agent can call.

  • coin_createLaunch on pump.fun: name, ticker, first buy and the creator fee split.
  • coin_launch_walletA practice wallet on devnet or the local chain, topped up with test SOL.
  • pumpfun_coin_infoPrice, market cap and how full the bonding curve is.
  • pumpfun_swapBuy or sell on the curve, and on PumpSwap once the coin graduates.
  • pumpfun_collect_feesClaim the creator fees into your wallet.
  • coin_retry_fee_splitSet the split again if it failed after a practice mint.
  • coin_set_take_profitSell the first buy at a target, watched by a DevGhouse job.
  • wallet_balanceRead the wallet right before anything moves money.

The DevGhouse guide tells agents to read the wallet and confirm with you before a launch or a trade. Mayhem mode and a buy-back share are options on the launch form for mainnet coins.

crypto.devghouse.com

The calllocal test chain
coin_create({
  chain: "solana", cluster: "localnet",
  name: "House", symbol: "DHOUSE",
  sol_lamports: 50000000,
  shareholders: [
    { address: "Hmdf…xDkZ", bps: 5000 },
    { address: "5Knr…Dqep", bps: 5000 },
  ],
})
sharing_config.status: "verified"
fee_shares_readback: 50% · 50%
The House coin page on crypto.devghouse.com: Solana and pump.fun badges, a Local chain test coin label and the creator fee split of 50% and 50% read from the chain
Local test chain, 12 October: House ($DHOUSE), launched by an agent through the DevGhouse MCP with 0.05 test SOL as the first buy and the split read back from the chain. Test coin, worth nothing.
YOUR SUBSCRIPTIONS

Your subscriptions.
Or ours, by the credit.

Link your Claude, Codex or Kimi subscription and your agents run on it, or let the built-in rail run a job for credits.

Claude

Your Claude subscription

Codex

Your Codex subscription

Kimi

Your Kimi account

Your DevGhouse

Your context. Your collaborators. Your work.

THE SAME JOB, YOUR CHOICE OF RAIL

Your subscription.
Or ours, by the credit.

  1. ClaudeRuns on your Claude Code subscription.
  2. CodexRuns on your Codex subscription.
  3. Built inDevGhouse runs it through OpenRouter, billed as credits.

Pick the rail in the model menu for each message. Every run shows the model it used and any credits it cost.

The same request answered on a Claude subscription, then on the built-in DevGhouse rail, where the run shows its credits
The real hub playing scripted example data. The replies and the 2 credits are example figures.

Subscription limits still apply. Other models can use metered credits; connected services and media jobs may have their own costs.

Connect your AI
WORKS WITH

Bring the tools you work with.

Your repositories, inboxes, planning tools and social accounts, in the context of your project. Connect the services you need and grant your agents the access their work requires.

AI, code and developer tools

Mail, search and productivity

Social channels

Available tools depend on the connection, account permissions and project. Explore the integrations ↗

Logos are trademarks of their owners; no endorsement implied.
ALSO ON DEVGHOUSE

One foundation.
Places to put it to work.

The same accounts, project context and connected tools carry into the apps we’re building and using ourselves.

APPS WE MADE

Built on DevGhouse.
Live today.

Standalone apps running on the same backend, each with its own people and its own work.