In Parallel

Registre officiel Productivité

The context layer for your company's AI: your team's decisions, meetings and plans, with sources.

Installation

Ajoutez ceci à la configuration MCP de votre client (Claude Desktop : claude_desktop_config.json ; Cursor : .cursor/mcp.json ; VS Code et les autres utilisent la même forme « mcpServers »).

Remote (streamable-http)
Claude Code
claude mcp add --transport http in-parallel 'https://www.in-parallel.ai/mcp'
Configuration JSON (Claude Desktop, Cursor, VS Code)
{
    "mcpServers": {
        "in-parallel": {
            "type": "http",
            "url": "https://www.in-parallel.ai/mcp"
        }
    }
}

Documentation

Tirée du README du projet sur GitHub, actualisée chaque semaine. Les liens ouvrent GitHub.

Sommaire
  1. In Parallel
  2. Get started
  3. Try it
  4. What the plugin does on your machine
  5. Need help?
  6. Publishing to the MCP registry
  7. Repository layout
  8. License

In Parallel

Bring your team's context into conversations with your AI assistant. Catch up on priorities, find relevant decisions, and share useful findings with your team.

Get started

You'll need an In Parallel account with access to your team's workspace.

Claude Code, Codex, or Cursor

Run the setup command. It asks which assistants to set up, then installs for your user:

npx -y github:in-parallel-oy/agent-plugin

Needs Node 20.12+. Claude Code and Codex also need their own CLI on your PATH. Cursor users should reload Cursor afterwards.

Later on, the same command takes doctor to check the installation and uninstall to remove it. Add --help for everything else.

Claude Desktop or ChatGPT

Open the app's plugin browser and install In Parallel from your team's plugin library. If it isn't listed, add https://github.com/in-parallel-oy/agent-plugin as a plugin source, or ask your administrator to make it available.

Claude Projects

Threads in a Claude project reach In Parallel through your claude.ai connector; the project conversation itself can't use In Parallel, only its threads can.

  1. Connect In Parallel under claude.ai → Customize → Connectors.
  2. Get the claude-projects ZIP from your administrator, or build it from a checkout of this repository with node scripts/package.js --client claude-projects. It holds the skill only, with no MCP server, hooks or scripts.
  3. Add it in Project settings → Plugins.

If your workspace offers it, Copy instructions for a Claude project on the workspace's What's set up page gives you project instructions to paste in.

Then

Sign in to In Parallel, choose the workspaces you want to connect, and start a new conversation with your assistant.

Try it

  • “What needs my attention this week?”
  • “Catch me up on [project].”
  • “Record this finding for my team: …”

What the plugin does on your machine

The plugin has two parts: the in-parallel skill and a connection to the In Parallel server. It runs no hooks or scripts on your machine and keeps no local files.

What is sent to In Parallel

Only what your assistant sends through the In Parallel connection (https://www.in-parallel.ai/mcp unless you chose another environment during setup). The skill asks your assistant to:

  • read team context, such as briefings, meetings, decisions and wiki pages
  • record findings you want to share with your team
Sign-in

You sign in through your assistant app's connection sign-in, which opens In Parallel in your browser. The app keeps the sign-in. The plugin's scripts never see, read or store a password or token.

The setup command

npx -y github:in-parallel-oy/agent-plugin downloads this repository and its two setup dependencies. It copies the plugin to ~/.in-parallel/setup/ for Claude Code and Codex, or ~/.cursor/plugins/local/in-parallel for Cursor, and registers it with the claude or codex command. It writes your chosen server URL into that copy.

Before it installs, it reads your existing MCP settings to find a conflicting In Parallel entry: ~/.claude.json (or the one in CLAUDE_CONFIG_DIR), .mcp.json in the current project, and ~/.cursor/mcp.json and .cursor/mcp.json. It never changes those files.

Setup and uninstall act on your user installation only. A copy installed for a single project is left in place; setup names it and the command that removes it.

doctor lists every location it checked (your user settings, the project root and the current directory, and each client's plugin registrations) as found, not found or conflicting, before it reports any error. For Claude Code it lists each In Parallel copy with its scope, the project it belongs to and its files, such as the setup copy for your user and one installed in another project. A conflicting entry says what it conflicts with and why, and a legacy entry named in-parallel comes with how to fix it. Changed installed files are named; when an installed .mcp.json gained an Authorization header, doctor says a credential header was added without printing it. It always shows the endpoint setup installed, even when a check fails. Without a setup copy, doctor lists any project copy setup did not install, says it can stay, and gives the command that removes it. It checks what is installed, not that it works: to check that, ask your agent "Check my In Parallel connection". doctor changes nothing.

Uninstall

If you used the setup command, run:

npx -y github:in-parallel-oy/agent-plugin uninstall

It removes only the files and registrations setup itself wrote, prints one line for each, and first backs the files up to ~/.in-parallel/setup/backups/. It stops without removing a file that changed since setup wrote it.

If you installed from a plugin browser, remove In Parallel there.

Need help?

Ask your assistant: “Check my In Parallel connection.”

If you can't install the plugin or access your workspace, ask your workspace administrator for help.

Publishing to the MCP registry

server.json lists the hosted In Parallel MCP server in the official MCP registry: the context layer for your company's AI, which gives Claude, ChatGPT, Copilot and other MCP clients your team's decisions, meetings and plans, with sources and scoped to each user's permissions. PulseMCP and Glama ingest new entries from the official registry; Smithery needs its own submission at smithery.ai.

Install mcp-publisher (brew install mcp-publisher) and check the file:

mcp-publisher validate

The name ai.in-parallel/in-parallel needs proof that you control in-parallel.ai. Both proofs use an Ed25519 key, which needs OpenSSL 3 (brew install openssl@3; macOS LibreSSL lacks Ed25519):

OPENSSL=/opt/homebrew/opt/openssl@3/bin/openssl
$OPENSSL genpkey -algorithm Ed25519 -out key.pem
PUBLIC_KEY="$($OPENSSL pkey -in key.pem -pubout -outform DER | tail -c 32 | base64)"
PRIVATE_KEY="$($OPENSSL pkey -in key.pem -noout -text | grep -A3 'priv:' | tail -n +2 | tr -d ' :\n')"

Then pick one:

  • DNS. Add a TXT record on the in-parallel.ai apex (not a subdomain) with the value v=MCPv1; k=ed25519; p=${PUBLIC_KEY}, wait for dig +short TXT in-parallel.ai to show it, then run mcp-publisher login dns --domain in-parallel.ai --private-key "$PRIVATE_KEY".
  • HTTP. Serve the single line v=MCPv1; k=ed25519; p=${PUBLIC_KEY} as text/plain at https://in-parallel.ai/.well-known/mcp-registry-auth. The registry fetches the apex and does not follow redirects, so the apex-to-www redirect must skip this path. Then run mcp-publisher login http --domain in-parallel.ai --private-key "$PRIVATE_KEY".

After logging in, publish:

mcp-publisher publish

Without domain access, rename the server to io.github.in-parallel-oy/in-parallel and log in with mcp-publisher login github as an Owner of the in-parallel-oy GitHub organization.

Keep key.pem out of the repo. Each published version is permanent, so bump the version and publish again when the listing should change. npm version <patch|minor|x.y.z> sets it in package.json, server.json and every plugin manifest, and commits the change.

Repository layout

Plugin directories and marketplaces install plugins/in-parallel/ only: the skill, the MCP connection, each client's manifest, LICENSE and NOTICE. The setup command, its package.json and lockfile and the build scripts stay at the repository root, outside the installed plugin. Keep plugins/in-parallel/LICENSE and plugins/in-parallel/NOTICE identical to the root copies; the tests check this.

License

Apache-2.0.

Questions fréquentes

Qu'est-ce que In Parallel ?

The context layer for your company's AI: your team's decisions, meetings and plans, with sources.

Comment installer In Parallel ?

Il fonctionne comme serveur distant. Dans Claude Code : claude mcp add --transport http in-parallel 'https://www.in-parallel.ai/mcp'. Pour les autres clients, collez la configuration JSON de cette page dans leurs réglages MCP.

Faut-il une clé API pour In Parallel ?

Son entrée au registre n'indique aucune clé ni secret obligatoire. Certaines fonctions peuvent tout de même demander un compte auprès du service auquel il se connecte.

In Parallel est-il open source ?

Oui. Son code est sur GitHub sous licence Apache-2.0.

Qui maintient In Parallel ?

Il est publié par in-parallel-oy. AI.tn le référence sans le maintenir ni le recommander.

Issu du registre MCP officiel. Être listé n'est pas une recommandation : vérifiez le code et les permissions d'un serveur avant de lui donner accès à vos données.