Voter
MCP server that gates AI changes through team voting, enabling collaborative approval workflows for any WithVibe environment.
README
Voter — a WithVibe plugin
Turn any WithVibe env into a team-voted workspace. When this plugin is enabled, the AI doesn't apply a prompt straight away — it opens a proposal the team votes on, and only an approved proposal gets built. Enable it on any env and that env becomes a vote-gated sandbox.
You (in chat): "add a dark-mode toggle"
│
▼
voter_propose_change ──────► Proposal #7 opens (10-min vote)
│ 👍 👎 team votes in the panel
voter_await_decision ◄─────── APPROVED ✅ / REJECTED ✗
│
▼ (only if approved)
agent makes the change ──► voter_record_change ──► shows up in History
It's content-agnostic — it knows nothing about what the env contains. Whatever the env is (an app, a doc, a 3D scene, a config), enabling Voter makes every change go through the team first. Multiple envs = multiple independent boards.
How the gate works (and why it's just a plugin)
WithVibe auto-injects an enabled plugin's MCP tools into the agent's turn. Voter uses that to gate prompts without any change to WithVibe core:
voter_propose_change(prompt)— opens a proposal and tells the agent to not touch anything yet.voter_await_decision(proposal_id)— long-polls the vote. The platform's MCP bridge caps a single tool call at 60s, so this returns after ~45s withSTILL_VOTINGand the agent simply calls it again until it's terminal.- On
APPROVED, the agent makes the change and callsvoter_record_change. OnREJECTED/EXPIREDit makes no change.
Meanwhile every teammate votes in the plugin's own UI panel — independent of the agent. The gate is agent-cooperative (enforced via tool instructions), which is the right trade-off for a friendly workflow; hardening it to be unbypassable is a later option that would need a small core hook.
MCP tools
| Tool | Purpose |
|---|---|
voter_propose_change |
Open a proposal for the requested change. Agent must call this instead of editing. |
voter_await_decision |
Long-poll the team's verdict (APPROVED / REJECTED / EXPIRED / STILL_VOTING). |
voter_record_change |
Record what shipped after applying an approved change. |
voter_report_failure |
Mark an approved change as failed (e.g. build broke). |
voter_status |
Show the current board: proposals in voting + recently resolved. |
State
Four tables in the plugin's isolated Postgres schema (shared-postgres storage):
settings (vote window + thresholds), proposal, vote, feed_event.
Resolution rule: a proposal passes when approvals reach the threshold, fails when rejections reach the threshold, and at the deadline the majority wins (ties expire). Defaults: 600s window, 1 approval / 1 rejection — tune them in the panel.
Architecture
manifest.yaml ──► WithVibe spawns one container per env (scope: env)
│
server.js (express)
├── /health platform health probe
├── /mcp agent's MCP endpoint ─► mcp.js (the gate)
├── /ui htmx voting panel ─► ui.js
└── /ui/version cheap freshness check (live updates)
│
db.js (pg) ─► shared-postgres, per-env schema
Build
docker build -t local/voter:0.1 .
For the marketplace this is published multi-arch as ghcr.io/withvibe/voter:<version>.
Install in WithVibe
- Workspace admin → Plugins → Install plugin (or Marketplace once listed).
- Paste manifest.yaml (or 1-click install from the catalog).
- Open any env → enable Voter → that env is now a vote-gated workspace.
Local development
docker build -t local/voter:0.1 .
docker run --rm -p 8080:8080 \
-e DATABASE_URL="postgres://user:pass@host.docker.internal:5432/withvibe_plugins" \
-e PGSCHEMA="voter_dev" \
local/voter:0.1
# open http://localhost:8080/ui — two browsers = two voters
Repository layout
manifest.yaml WithVibe plugin manifest (install input)
Dockerfile runtime image
package.json deps (express, pg, @modelcontextprotocol/sdk, zod)
server.js express entry — HTTP + MCP routes + vote handling
db.js pg pool, schema, proposal/vote/resolution helpers
mcp.js the gate: propose / await / record MCP tools
ui.js htmx voting panel + dark theme
License
Apache License 2.0. See NOTICE for attribution.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.