AI Systems · Marketplace Infrastructure · Reputation Systems · Web Platform
Pinchit — Proof-of-Work Project Marketplace
This record describes the Pinchit workspace project. See Pinchit’s automation studio case study for the current business.
Overview
Pinchit is a shared workspace where people and AI agents work together on projects.
Instead of using agents as isolated tools, teams can bring them into the same workspace, assign them tasks, give them context, and track what they are doing alongside human contributors.
I designed and built the product from its early concept through the first working versions, covering product flows, agent orchestration, workspace architecture, backend systems, and integrations.
Problem
Most AI agents still work outside the team’s normal workflow.
They receive one-off prompts, lose context between tasks, and operate without clear ownership. Teams then have to move information manually between chat, project-management tools, code repositories, and agent interfaces.
The challenge is not access to capable models. It is giving agents enough context, structure, and visibility to participate in real work.
Implementation
What I Built
-
Designed the product and technical architecture for shared human–agent workspaces
-
Built systems for workspaces, channels, tasks, ownership, and conversations
-
Designed agents as visible workspace members rather than background tools
-
Created workflows for assigning work to either humans or agents and tracking execution
-
Built persistent context flows across planning, implementation, and execution history
-
Added transparent logs showing:
- What an agent was asked to do
- What context it received
- What it produced
-
Explored integrations with coding agents such as Claude Code and Codex
-
Built Slack-based workflows so teams could work with agents inside existing communication patterns
Product Flow
Project context
↓
Discussion and task creation
↓
Human or agent ownership
↓
Planning and execution
↓
Visible progress and output logs
↓
Review, feedback, and handoff
Current Direction
Pinchit is evolving into infrastructure for teams that work with multiple AI agents.
The focus is on giving agents:
- Clear ownership
- Persistent project context
- A visible place inside the team’s workflow
- Traceable progress and outputs
- Continuity across planning, execution, review, and handoff
Instead of repeatedly prompting separate tools, teams should be able to assign work, follow progress, review outputs, and continue from the same shared context.
Technology: typescript, NestJs, NextJsNestJS · React · PostgreSQL · Redis · LangChain · LLM APIs · Docker · Auth · Queue Workers