ONE DURABLE RECORD
EVERY AGENT READS/WRITES
Scout is the one record every coding agent reads at session start and writes to as it works. A durable spine for your AI fleet.
- What is Scout?
- Scout is the one durable record of what the team intended, changed, and decided. One record. Every agent reads it. Every agent writes it. Only what’s current reaches the model.
- How do coding agents share one record?
- Scout maintains one record. As agents work, they write their intent. When they start, they read the log. Every entry is kept. The model never reads the whole history. Scout projects what is currently true: newest entry wins, contested entries are flagged, unconfirmed entries expire.
- How is Scout different from a Cursor rules file, CLAUDE.md, or AGENTS.md?
- Those files are a static paste per tool. They drift when the rule changes or the team switches models. Scout is not a better markdown file. Connecting MCP is the setup. The agent reads the team’s rules through the session bundle. Change the record once. The next session gets it.
- What does Scout cost?
- Team is $14.99 per human developer per month. Fifteen days, every seat, card up front, first charge on day 15. Under $1M raised or under $250K annual revenue, the startup program is free for 12 months: 5 developers and 5,000 entries a month. There is no free plan.
- Our coding agents don’t share context. What fixes that?
- Scout is the one record every coding agent reads at session start and writes to as it works. One record. Every agent reads it. Every agent writes it. Only what’s current reaches the model.
- Each tool remembers only itself. Cursor doesn’t see Claude, and Claude doesn’t see Cursor.
- Cursor remembers Cursor. Copilot remembers Copilot. Claude remembers Claude. Scout is one MCP record. The record survives switching tools, for every agent that speaks MCP.
- Our CLAUDE.md and Cursor rules keep going stale.
- Those files are a static paste per tool. They drift when the rule changes or the team switches models. Scout is not a better markdown file. Change the record once. The next session gets it.
- We re-brief the agent at the start of every session.
- Scout is the one durable record of what the team intended, changed, and decided. Kept current so every agent and the person running the team sees what is true now. No re-briefing. No drift.
- The model is reading the whole history and the session gets expensive.
- Every entry is kept. The model never reads the whole history. Scout projects what is currently true: newest entry wins, contested entries are flagged, unconfirmed entries expire. Scout is not a token compressor or summarizer.