
Pabs
io.github.pdunovitsPublished by GitHub account pdunovitsv1.0.0Updated Oct 24, 2026
The decision layer between an idea and a digital product that works.
Overview
PABS exposes a product strategy plan — brief, hypotheses, PRD, competitor analysis and pulse — to AI coding tools, and feeds build observations back for…
- What you can do
- Read strategy: fetch the full context in one call, or separately the brief, product strategy, competitor analysis, PRD, documents, decisions, hypotheses, action plan, pulse, learnings and pulse schedule
- Ask the strategy agent a question and get alignment scores, recommendations and considerations
- Report real-world observations into the strategy feedback loop and trigger a fresh strategic pulse
- Review pending learnings by approving, rejecting or modifying them inline
- Set the pulse cadence by frequency, day, hour and timezone
- When to use it
- Useful when an AI coding assistant should build against a defined product strategy instead of guessing, and when you want build-time observations and drift reported back into a strategy platform for review without leaving the coding session.
- Requirements
- A remote Streamable HTTP endpoint at pabs.co.nz; an active PABS account with at least one saved plan containing a brief; a per-plan token generated in the PABS Connect tab, supplied in the server URL path, as a Bearer token in the Authorization header, or in the X-PABS-Token header.
Installation
In SourceWeft
- Open Pabs in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.
Other MCP clients
Follow the launch instructions in the repository.
README
PABS MCP Server
Give your AI coding tools live access to your product strategy.
PABS is a product strategy platform that turns a brief into a full strategy — hypotheses, competitor analysis, PRD, and a living pulse of your venture's health. The MCP server exposes that strategy to any MCP-compatible AI tool so it can build with context, not guesses.
When your AI builder knows the target audience, design hypotheses, and current strategic verdict, it makes better decisions. When it spots a gap between what was planned and what was built, it can report it back — and PABS queues it for your review.
Connection
Transport: Streamable HTTP (MCP 2025 spec)
Server URL: https://pabs.co.nz/api/mcp/{your-token}
Authentication: Bearer token (one per plan, generated in the PABS app)
Get your token
- Open your plan in PABS
- Go to the Connect tab (or click the plug icon)
- Click Generate Token
- Copy the server URL — it includes your token
Connect to your AI tool
Claude Desktop
Add to claude_desktop_config.json:
Claude Code
Cursor
Add to .cursor/mcp.json in your project (or global settings):
Windsurf
Add to ~/.codeium/windsurf/mcp_config.json:
Replit Agent
- Go to replit.com/integrations → MCP Servers
- Click Add MCP Server
- Enter a name (e.g.
PABS - My App) - Server URL:
https://pabs.co.nz/api/mcp/YOUR_TOKEN - Click Test & Save
Any MCP-compatible client
The server speaks the MCP 2025 Streamable HTTP transport. Send POST requests to the server URL with a Bearer token in the Authorization header, or embed the token directly in the URL path (as shown above).
Tools
Read your strategy
Act on strategy
Schedule
The feedback loop
PABS is designed for a two-way relationship between AI builders and the strategy layer.
Outbound (strategy to builder): The builder calls get_full_context at the start of a session to load the current strategy, hypotheses, target audience, and decisions. It builds with that context.
Inbound (builder to strategy): As the builder works, it can call ask_pabs_agent to check whether a proposed feature aligns. If it observes something real — a drop-off, a pattern, a change in scope — it calls report_learning. PABS evaluates the learning against the strategy, proposes an adaptation, and notifies the founder.
Human in the loop: Strategy only updates after the founder reviews and approves a learning. get_pending_learnings and decide_learning let that review happen inline, inside the builder session.
Authentication
Each plan has a unique token. Tokens do not expire by default. You can revoke and regenerate a token at any time from the PABS Connect tab.
The server URL with the token embedded (/api/mcp/{token}) is the recommended format — it lets AI clients distinguish between multiple plans as separate servers.
Alternatively, pass the token in the Authorization header as Bearer {token}, or in the X-PABS-Token header.
Security
- Each token is scoped to a single plan. A token cannot access data from any other plan.
- The server validates the token on every request before any data is returned.
- All requests go over HTTPS.
Requirements
- An active PABS account
- At least one saved plan with a brief
- A token generated from the Connect tab
Links
License
This repository documents the hosted PABS MCP server. The server is proprietary and operated by PABS. Client configuration examples in examples/ are released under MIT.
Source: README.md at commit b33cec6
Tools
0Version history
1- v1.0.0LatestSep 16, 2026
Score
40% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 daysRequired25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declaredRequired10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts
Declare a license in the repository or the registry entry.