About bridgeapp
The bridgeapp plugin connects coding agents to a BridgeApp workspace over MCP, allowing access to tasks, chats, threads, pages, and projects under the signed-in account's permissions. It also includes skills for task work, context gathering, and investigation.
Key features
- 01Read and act on tasks and projects
- 02Access chats, threads, and pages
- 03Skills for tasks and workspace context
- 04Shared server and skills across clients
- 05Workspace access follows account permissions
Use cases
- 01Work from task plans, comments, and blockers
- 02Read shared messages and agent transcripts
- 03Trace related context from tasks, messages, or pages
- 04Investigate issues spanning a repository and BridgeApp
MCP support
The plugin connects to BridgeApp's MCP service, hosted at https://mcp.bridgeapp.ai/mcp and authorized with OAuth. Other MCP clients supporting Streamable HTTP can connect; self-hosted workspaces require registering their own endpoint.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.bridgeapp.ai/mcpclaude mcp add --transport http ai-bridgeapp-bridgeapp https://mcp.bridgeapp.ai/mcp{"mcpServers":{"ai-bridgeapp-bridgeapp":{"type":"http","url":"https://mcp.bridgeapp.ai/mcp"}}}Open in Cursor{"servers":{"ai-bridgeapp-bridgeapp":{"type":"http","url":"https://mcp.bridgeapp.ai/mcp"}}}Open in VS Code{"mcpServers":{"ai-bridgeapp-bridgeapp":{"type":"http","url":"https://mcp.bridgeapp.ai/mcp"}}}{"mcpServers":{"ai-bridgeapp-bridgeapp":{"serverUrl":"https://mcp.bridgeapp.ai/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEBridgeApp plugins
Official plugins that connect a coding agent to a BridgeApp workspace over MCP, so it can read and act on tasks, chats, threads, pages, and projects under the signed-in account's own permissions.
Claude and Claude Code
BridgeApp is in the Claude directory. Open the BridgeApp plugin and install it — or, in Claude on the web, on desktop or in Cowork, open Settings → Plugins → Discover, search for BridgeApp and install it. Then open the plugin's Connectors tab, choose BridgeApp and click Connect to sign in.
A plugin added to a Claude account reaches Claude Code on its own, as
bridgeapp@synced, whenever Claude Code signs in with the same account
(Claude Code 2.1.273 or later). Restart Claude Code or type /reload-plugins
to load it, then sign in from /mcp if BridgeApp asks for it.
Claude Code without a Claude account — an API key, Amazon Bedrock or Google Vertex AI — installs the plugin from this repository's marketplace instead. Do not do this with a Claude account as well: the directory copy already syncs, and a second copy takes its place.
claude plugin marketplace add MathAndMagic/bridgeapp-plugin
claude plugin install bridgeapp@bridgeapp
Then sign in — this needs an interactive terminal, so run it yourself rather
than asking an agent to. Claude Code addresses a plugin-provided server as
plugin:<plugin>:<server>, unless a bridgeapp server was already registered
by hand — then it keeps that entry, adds no plugin copy, and the bare name is
the one to sign in to:
claude mcp login plugin:bridgeapp:bridgeapp || claude mcp login bridgeapp
Codex and ChatGPT
The same plugin installs into Codex — MCP server and skills together:
codex plugin marketplace add MathAndMagic/bridgeapp-plugin
codex plugin add bridgeapp@bridgeapp
Codex asks you to sign in to BridgeApp the first time a tool is used. The
ChatGPT desktop app shares ~/.codex with Codex, so the install above serves
both. To install from the app instead, open Plugins, choose Add
marketplace and then Add from a repository, enter
MathAndMagic/bridgeapp-plugin, and install BridgeApp from that marketplace.
The manifest lives in .codex-plugin/plugin.json and the marketplace in
.agents/plugins/marketplace.json; both read the same skills/ and .mcp.json
as the Claude Code and Cursor manifests. The manifest also carries the OpenAI
directory listing and the review metadata (test cases, commerce declaration,
release notes) under extensions["com.openai"]. Reviewer credentials and the
demo video URL are entered in the OpenAI portal, never in this repository.
Other agents
Point any MCP client that speaks streamable HTTP with OAuth at:
https://mcp.bridgeapp.ai/mcp
It discovers the authorization server and the bridgeapp:mcp scope on its own
from https://mcp.bridgeapp.ai/.well-known/oauth-protected-resource.
Per-client instructions, including the ones an agent can follow on its own, live
in agent-setup.md, published at
https://bridgeapp.ai/agent-setup.md and at /agent-setup.md on every
workspace host.
That file is the master copy. The same file ships in bridgeapp-web
(public/agent-setup.md, serving every workspace host) and in bridge-website
(public/agent-setup.md, serving the apex domain) — edit it here and copy it to
both in the same piece of work, byte-identical.
Self-hosted workspaces
Every URL above belongs to the hosted service. BridgeApp also runs on customers'
own infrastructure, under domains of their choosing, and the plugin ships the
hosted endpoint in .mcp.json — deliberately
as a plain value: Claude Code substitutes plugin settings unreliably across
versions, and a server that silently fails to register is worse than a fixed
URL. On a self-hosted workspace, install the plugin first anyway — the skills
apply to any workspace — and then register the workspace's own endpoint, shown
on its Agents → Connect Apps page, next to it.
Claude, with a Claude account: open Settings → Connectors, click +, choose Add custom connector, paste the workspace's endpoint and connect it. Leave the BridgeApp connector that came with the plugin unconnected. Claude Code signed in with the same account picks up both.
Claude Code without a Claude account:
claude mcp add --scope user --transport http bridgeapp https://mcp.example.com/mcp
claude mcp login bridgeapp
Codex and the ChatGPT desktop app, under a name of its own so it does not
replace the plugin's bridgeapp entry (add starts the sign-in itself):
codex mcp add bridgeapp-workspace --url https://mcp.example.com/mcp
The plugin's own hosted server stays unauthenticated — do not sign in to it —
and can be ignored or removed. Links back into the workspace come from the
server's build_bridgeapp_link tool, so they carry the self-hosted domain on
their own.
What is in here
plugins/bridgeapp/
.claude-plugin/plugin.json Claude Code manifest
.cursor-plugin/plugin.json Cursor manifest
.codex-plugin/plugin.json Codex / ChatGPT manifest (OpenAI directory listing and review metadata)
.mcp.json the hosted BridgeApp MCP server
skills/ how to work with a BridgeApp workspace
One plugin serves every client: the manifests differ, the MCP server and the skills are shared.
Skills
- bridgeapp-task — working from a task: its plan, comments, subtasks and blockers, plus the project brief, which is the only place the repository and the team's flow are recorded.
- bridgeapp-thread — reading a shared message or thread to the end, including the agent transcripts inside it. This is what fires when someone shares a message from BridgeApp with your agent.
- bridgeapp-context — walking from a task, message, or page to everything it hangs off, and where to stop.
- bridgeapp-investigation — diagnosing a bug when the trail runs through both the repository and BridgeApp, or finding out what was decided about something and by whom. Read-only by construction.
- bridgeapp-task-creation — turning a request into a well-formed task, including the duplicate check that comes first.
- bridgeapp-spec — writing a spec from the decisions already taken, asking the author about contradictions and gaps.
Because the skills carry the how, a shared link only has to carry the what: the Share action in BridgeApp sends a pointer to the message, not a copy of the conversation and not a list of instructions.
Verifying a connection
Call list_projects. An empty list is a pass: it means the connection works and
the account has no projects yet.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.