Mercury

Overview

Mercury appears in this corpus as both a fintech product and an internal operating environment for advanced AI-assisted knowledge work. In the newsletter mentions, Mercury is referenced in two main ways: first, as a company whose product APIs and connectors can be integrated into AI workflows; and second, as an organization whose internal documents, specs, queries, Slack threads, and transcripts were indexed into a Claude Code knowledge base. For AI Product Managers, that makes Mercury notable not just as a software company, but as a concrete example of how modern product teams are turning years of institutional knowledge into searchable, AI-native systems.

The strongest signal is the reported ingestion of nearly 5 million words from five years of Mercury PM work into a local QMD-indexed Claude Code setup, creating a "second brain" that reportedly doubled productivity. Separately, Mercury’s API and MCP-related mentions suggest a product stack designed for secure integrations and practical agent use cases. Together, these references position Mercury as a useful case study in enterprise knowledge management, AI copilots for PM work, and the importance of robust underlying APIs before adding agent-facing abstractions.

Key Developments

• 2026-02-08: Tal Raviv reportedly gave Opus 4.5 read-only access to his Mercury bank account through Mercury’s MCP connector, described as an official Anthropic app with quick OAuth, to help diagnose a tax shortfall.
• 2026-04-08: Peter Yang referenced wiring Mercury, Google Workspace, and other APIs into his OpenClaw AI agent to automate the first 80% of documents, slides, and analytics before manual refinement.
• 2026-04-23: Ryan Wiggs explained that Mercury prioritizes building robust APIs before MCPs, and described ingesting 5 million words from five years of PM work into Claude Code via QMD search to create a productivity-enhancing "second brain."
• 2026-04-28: Ryan Wiggins was noted as having built a local QMD-indexed Claude Code knowledge base from nearly 5 million words of Mercury strategy docs, specs, queries, Slack threads, and transcripts.

Relevance to AI PMs

1. A blueprint for AI-native knowledge management: Mercury’s reported Claude Code knowledge base shows how PM teams can transform fragmented institutional knowledge—strategy docs, product specs, Slack conversations, transcripts, and queries—into a searchable AI context layer. AI PMs can apply this pattern to reduce repeated research, speed onboarding, and improve decision continuity.

2. A practical lesson in API-first agent design: The mention that Mercury builds robust APIs before MCPs is tactically important. AI PMs evaluating agent experiences should treat clean APIs, permissions, and structured access patterns as the foundation; agent protocols and connectors become more useful when the core product surface is already reliable and well-scoped.

3. An example of secure financial and workflow integration: Mercury’s MCP connector and OpenClaw/API integrations highlight real-world AI use cases where sensitive business systems can be connected to models with controlled access. For AI PMs, this reinforces the need to design around read-only modes, OAuth flows, auditability, and narrow permissions when bringing AI into operational workflows.

Related

• google-workspace: Mentioned alongside Mercury as part of an API stack connected into AI agents for automating docs, slides, and analytics.
• openclaw: Peter Yang’s AI agent, which was wired into Mercury and other systems to automate substantial portions of knowledge work.
• peter-yang: Referenced Mercury in the context of workflow automation and API-connected AI agents.
• tal-raviv: Used Mercury’s MCP connector with Opus 4.5 for a financial diagnostic use case.
• opus-45: The model given read-only access to Mercury account data in the tax-shortfall example.
• mcp: Central to Mercury’s connector story and relevant to how AI systems access tools and data sources.
• anthropic: Mentioned because Mercury’s MCP connector was described as an official Anthropic app.
• ryan-wiggs / ryan-wiggins: Credited with describing or building the Mercury knowledge base from years of PM materials.
• claude-code: The environment used to build and query the Mercury knowledge base.
• qmd: The indexing/search layer used to make Mercury’s large internal corpus accessible inside Claude Code.