Decisions
The what. And the why.
Keep architecture choices and constraints close to the work, including the reasoning behind them.
example · decision
- database
- PostgreSQL
- reason
- relational reporting
- constraint
- no schema changes this sprint
MCP CONTEXT LAYER FOR AI TOOLS
Switch your AI. Keep your context.
The decisions you made in Claude belong in Cursor, too. Save project knowledge to Webcraw and ask your next AI tool to pick it up.
Works with
- Claude
- Cursor
- Claude Code
- ChatGPT
HOW IT WORKS
Save. Switch. Pick up.
You already explained it. Then you switched tools. Webcraw keeps the decisions, preferences, and handoffs you choose to save, so the next conversation starts where the last one ended.
- 01
Connect your tools
Add Webcraw through MCP. Authorize each AI tool for the same workspace.
Claude + Cursor → Webcraw - 02
Save what matters
Ask your AI to store a decision, preference, or brief. Check the saved result.
“Save our Atlas project context.” - 03
Ask for it elsewhere
Start in your next tool. Read the project context before continuing the work.
“Read Atlas from Webcraw.”
WHAT YOU KEEP
Keep what moves the project forward.
Small, useful pieces of knowledge, ready for the next conversation, plus the tools to put them to work.
Project briefs
A place to pick up.
Save the state of the work so your next session starts with the open questions and next steps.
example · brief
- done
- signup implementation
- open
- email provider choice
- next
- verify mobile signup
Preferences
Your way of working.
Carry writing style, naming conventions, and review expectations from one assistant to another.
Sandboxes
Cloud sandboxes
Run Python and JavaScript in the cloud and keep reusable scripts alongside project knowledge.
Pro
Scheduled scripts
Schedule recurring tasks and put your saved scripts to work.
Secrets
Environment variables
Manage environment variables and secrets, and link them to scripts for automatic injection.
QUICKSTART
Make your first handoff.
- 1
Connect the MCP server
Add the Webcraw server to Claude, Cursor or Claude Code.
- 2
Save from one tool
Ask it to write your project brief as a ROOT knowledge node.
- 3
Pick up in another
Open a fresh conversation elsewhere and read the same node.
Cursor MCP config
{
"mcpServers": {
"webcraw": { "url": "https://mcp.webcraw.ai/http" }
}
}Example prompt
Use Webcraw to save our Atlas project context. First check whether a knowledge node named "Atlas project context" exists. If it exists, read it and preserve unrelated content. Otherwise create it with the ROOT label so it is discoverable.
Save these decisions:
- Database: PostgreSQL.
- Timestamps: store in UTC.
- Deployment: require review before deployment.PRICING
Try the whole handoff. On the free plan.
Save, retrieve, and update knowledge. Upgrade when you need more capacity.
Free
START HERE$0/mo
Make your first handoff. Keep a project’s context across tools.
Start with shared context- Shared context through MCP
- Up to 500 knowledge nodes
- 150 MCP requests per 5 hours
- 10 reusable scripts
- Built-in AI chat · 15 messages per 5 hours
- Or use your Claude or Codex subscription via MCP
- Community support
Pro
$5/mo
More room for your knowledge, with tools for recurring work.
Create an account- Shared context through MCP
- Higher knowledge & script limits
- 1,000 MCP requests per 5 hours
- Scheduled tasks
- Built-in AI chat
- Or use your Claude or Codex subscription via MCP
- Priority sandbox execution & email support
The message limit applies to Webcraw’s built-in chat. You can also connect Webcraw to Claude or Codex using your own subscription. Webcraw MCP request limits and your AI client’s plan and connector requirements still apply.
Need self-hosting or custom controls? Discuss Enterprise requirements
FAQ
Know what you're connecting.
How Webcraw handles your context, your tools, and your data.
Does Webcraw automatically sync my conversations?
No. You ask your connected AI tool to save selected knowledge to Webcraw and to retrieve it when needed. Full chat histories are not imported, and a new conversation does not automatically load your saved context.
What should I save?
Start with a small project brief: important decisions, constraints, working preferences, and next steps. Give it a clear project name. Ask your AI to update the existing brief as decisions change.
How does another tool find my context?
Connect both tools to the same Webcraw account and workspace. Ask the second tool to read the project by name. Webcraw’s overview tool helps your AI discover saved knowledge. The quickstart includes exact prompts and troubleshooting.
Where is the knowledge stored?
Knowledge saved through the hosted service is stored in your Webcraw workspace. Connected clients can read or change it through MCP within their authorized permissions. Review the knowledge you want to make available before saving it.
Can I try shared context on the free plan?
Yes. The free plan includes shared context through MCP, up to 500 knowledge nodes, and 150 MCP requests per five hours. The Claude-to-Cursor tutorial fits within these limits and does not need scheduled tasks.
What happens when a decision changes?
Ask your AI to read and update the existing project knowledge, preserving unrelated details. In the other tool, ask for a fresh read. A conversation may still contain an older copy until you retrieve the updated version.
Your next conversation can start further along.
One project brief. Two connected tools. Try your first real handoff.