proxy.ing Routes
proxy.ing separates web access, app connections, and host APIs:
- Web traffic for the shared public chat client.
- OpenAI-compatible
/v1/*traffic for browser-safe API use. - Authenticated private tunnel traffic for Grand Central client, inference, and MCP routes.
- Authorization metadata and OAuth routes for compatible connected apps.
An address identifies the host. Permission comes from the approved device or the connection the owner approved. Start with Connect your app when building a third-party client; do not assume that access to an address opens all of the routes below.
Public Route Map
| Route | Purpose | CORS |
|---|---|---|
https://username.proxy.ing/ |
Shared web chat client | Site default |
https://username.proxy.ing/v1/* |
OpenAI-compatible route on the user's tunnel origin | Browser CORS enabled |
https://username.proxy.ing/client/* |
Grand Central client API | No broad browser CORS |
https://username.proxy.ing/inference/* |
Grand Central inference API | No broad browser CORS |
https://username.proxy.ing/mcp/proxy |
Proxy MCP server | No broad browser CORS |
https://username.proxy.ing/mcp/party |
Party MCP server | No broad browser CORS |
https://username.proxy.ing/.well-known/oauth-authorization-server |
Authorization metadata on a compatible host | Browser CORS enabled |
https://username.proxy.ing/oauth/* |
App registration and owner-approved connection flow | Browser CORS enabled |
Unknown usernames and reserved subdomains redirect to https://proxy.ing.
Tunnel Ingress
Proxy creates a Cloudflare Tunnel configuration for the user's address. The private prefixes route to Grand Central on the device:
| Prefix | Local service |
|---|---|
/mcp/* |
http://localhost:51711 |
/client/* |
http://localhost:51711 |
/inference/* |
http://localhost:51711 |
| everything else | The configured local web/chat origin |
The fallback local origin can use a different port, but the Grand Central
client, inference, and MCP prefixes route to port 51711.
Grand Central Client Routes
When accessed through proxy.ing, these routes live under
https://username.proxy.ing/client/v1.
| Route | Purpose |
|---|---|
GET /timeline/entries |
Timeline entries for the app surface |
GET /timeline/entries/:id |
One timeline entry |
GET /content/:reference |
Referenced content blobs |
GET /events |
App event stream |
GET /moves |
Moves for review |
POST /moves/:id/resolve |
Resolve a Move with proof from an approved owner device |
GET /agents |
Available agents |
GET /agents/:id |
Agent detail |
GET /agents/:id/loadouts |
Agent loadouts |
POST /agents/:id/loadouts |
Create a loadout |
PATCH /loadouts/:id |
Update a loadout |
DELETE /loadouts/:id |
Delete a loadout |
POST /loadouts/:id/duplicate |
Duplicate a loadout |
POST /loadouts/:id/activate |
Activate a loadout |
POST/PATCH/DELETE /loadouts/:id/skills/:skill_id |
Equip, update, or remove a skill |
POST/DELETE /loadouts/:id/integrations/:integration_id |
Equip or remove an integration |
GET /catalog/harnesses |
Harness catalog |
GET /catalog/skills |
Skill catalog |
GET /catalog/integrations |
Integration catalog |
GET /catalog/models |
Model catalog |
GET /catalog/providers |
Provider catalog |
GET/POST /threads |
List or create threads |
GET /threads/:id |
Thread detail |
GET /threads/:id/roster |
Thread roster |
POST /threads/:id/participants |
Add a participant |
GET /threads/:id/snapshot |
Thread snapshot |
GET /threads/:id/events |
Thread event stream |
GET/POST /threads/:id/messages |
List or send messages |
POST /approvals/:id/respond |
Respond to an approval request |
Grand Central Inference Routes
When accessed through proxy.ing, these routes live under
https://username.proxy.ing/inference/v1.
| Route | Purpose |
|---|---|
POST /responses |
OpenAI Responses-style inference |
POST /chat/completions |
OpenAI-compatible chat completions |
POST /completions |
OpenAI-compatible text completions |
POST /embeddings |
Embeddings |
GET /models |
Model listing |
GET /models/:model_id/status |
Model readiness |
For browser clients that need broad CORS, use the top-level /v1/* route class.
For app and device clients that carry the expected bearer token, use
/inference/v1/*.
MCP Routes
proxy.ing currently documents two MCP endpoints:
| Route | Purpose |
|---|---|
/mcp/proxy |
Proxy MCP tools |
/mcp/party |
Party MCP tools |
These are private tunnel routes. Treat them as authenticated device routes, not public browser APIs.