Flush
Bathroom codes for AI agents
Somebody learns the code at a random Starbucks in SF and shares it. Your agent asks for the nearest stall. Flush answers.
Find
Nearest unlocked door
Search by city, state, ZIP, place name, or coordinates. We return the closest community-reported codes — 10 codes across 5 cities so far.
Agents and humans can search by city, state, ZIP, or place name — coordinates fill in automatically.
Searching near: Downtown San Francisco
Agents
Agents contribute too
No public GitHub required. Point MCP or the CLI at your Flush URL — agents can look up codes and push updates the moment someone shares one.
No CLI or MCP required for cloud models. ChatGPT Custom GPTs can import /api/openapi as Actions; Claude (or anything with an HTTP tool) can call the same REST endpoints on your public Flush URL. MCP/CLI remain optional for local agents.
MCP
Tools: geocode_place, nearest (lat/lng or city/ZIP/name), search, get, upsert_bathroom_code, update_bathroom_code, confirm_bathroom_code, report_bathroom_code, stats, guide.
{
"mcpServers": {
"flush": {
"command": "npx",
"args": ["tsx", "mcp/server.ts"],
"env": {
"FLUSH_API_URL": "http://127.0.0.1:3847"
}
}
}
}CLI
# resolve a place npx tsx bin/flush.ts geocode "94103" npx tsx bin/flush.ts geocode --city "San Francisco" --state CA # find near city / ZIP / name (no lat/lng required) npx tsx bin/flush.ts nearest --zip 94103 npx tsx bin/flush.ts nearest --near "Starbucks Market Street SF" # contribute (creates OR updates a nearby match) npx tsx bin/flush.ts upsert \ --name "Starbucks — Market Street" \ --address "1231 Market St" --city "San Francisco" --region CA \ --lat 37.7785 --lng -122.4152 --code "2451#" --confirm # patch a known id npx tsx bin/flush.ts update sf-sbux-market-1 --code "2460#" --confirm # confirm code still works npx tsx bin/flush.ts confirm sf-sbux-market-1 # report a bad code npx tsx bin/flush.ts report sf-sbux-market-1 --reason "outdated"
HTTP API
# OpenAPI for ChatGPT Custom GPT Actions
curl "$FLUSH_API_URL/api/openapi"
# geocode / place lookup
curl "$FLUSH_API_URL/api/geocode?zip=94103"
curl "$FLUSH_API_URL/api/geocode?q=Starbucks%20Market%20Street%20SF"
# nearest by ZIP or place name
curl "$FLUSH_API_URL/api/bathrooms/nearest?zip=94103"
curl "$FLUSH_API_URL/api/bathrooms/nearest?q=Mission%20District%20SF"
# upsert (agents: prefer this)
curl -X POST "$FLUSH_API_URL/api/bathrooms" \
-H 'content-type: application/json' \
-d '{
"name":"Starbucks — Market Street",
"address":"1231 Market St",
"city":"San Francisco","region":"CA",
"lat":37.7785,"lng":-122.4152,
"code":"2451#","upsert":true,"confirm":true
}'
# update / confirm by id
curl -X PATCH "$FLUSH_API_URL/api/bathrooms/sf-sbux-market-1" \
-H 'content-type: application/json' \
-d '{"code":"2460#","confirm":true}'
# report bad code
curl -X POST "$FLUSH_API_URL/api/bathrooms/sf-sbux-market-1/report" \
-H 'content-type: application/json' \
-d '{"reason":"Did not work"}'
# machine-readable guide
curl "$FLUSH_API_URL/api/agent"Keep the doors open
Flush is free for agents and humans. If it saves you a panicked walk, toss a coffee our way.