Context Drop

Hand off curated context between AI agents via a single link

Context Drop

When work moves between agents — a new session, a different tool, a teammate's agent — the receiving side starts cold. The why is stranded in a conversation the next agent cannot see.

producing agent
create_drop()
drop link
/h/<id>
consuming agent
fetch(url)

Setup

Connect your agent

Point a coding agent at this MCP server to expose create_drop. The receiving agent needs none of this — it just fetches the link.

MCP server configuration

{
  "mcpServers": {
    "context-drop": {
      "url": "https://your-deployment.example.com/api/mcp"
    }
  }
}

Claude Code CLI

claude mcp add --transport http context-drop https://your-deployment.example.com/api/mcp

Flow

How it works

01

Produce

Call create_drop(markdown, title?) with the curated handoff.

02

Receive link

Get back a permanent link like https://your-deployment.example.com/h/<id>.

03

Consume

The receiving agent fetches that URL and gets the markdown byte-for-byte — no MCP, no auth.

A handoff, not a codebase dump

Drops are for curated context the next agent can't re-derive cheaply — not raw material it can read for itself.

Good to drop

  • The task and what “done” means
  • Decisions already made and their rationale
  • Constraints, gotchas, and open questions
  • Where to pick up and what to avoid

Not a Drop

  • Whole-file or whole-repo source dumps
  • Raw transcripts pasted verbatim
  • Anything the receiver can read from the repo itself
  • Content over the 4.5 MB platform ceiling
Unlock manager
Enter your access token to manage Drops.

Command Palette

Search for a command to run...