Use our backend with the coding tool you already like.
Cline, OpenCode, Continue, Hermes, and OpenAI-compatible IDE clients can connect directly to Platform Core. Forge/Kiln is the login, key, receipt, docs, and tips-only support surface.
https://platform-core-api.tashdid.workers.dev/v1https://platform-core-api.tashdid.workers.dev/api/v1/mcphttps://platform-core-api.tashdid.workers.dev/api/v1/public/ide-backend/statushttps://platform-core-api.tashdid.workers.dev/api/v1/public/ide-backend/claim-packYou only paste a base URL and virtual API key into your IDE. Tool execution, MCP, workspaces, receipts, and evidence stay behind scoped permissions.
Auth users understand
Log in to Forge/Kiln, create a scoped virtual IDE key, copy the config, and revoke or rotate the key whenever needed.
At-cost receipts
Model, embedding, runtime, and tool surfaces produce usage receipts. Tips are optional dashboard CTAs, not hidden IDE protocol fees.
Tool-safe backend
The `/v1` route proposes model/tool output. Shell, file, MCP, daemon, and sandbox work stay permissioned, scoped, auditable, and deny-by-default.
What is safe to claim today?
The product surface is implemented for external IDE clients. Stronger wording like "live-tested in every desktop client" is intentionally gated behind recorded evidence rows, receipts, and screenshots.
- Backend setup flow, key management, receipts, runtime bridge, and evidence matrix exist.
- Users can configure supported clients with the /v1 base URL and a scoped key.
- Public all-client live-tested claims require the evidence matrix to be complete.
Client proof matrix
model chat, streaming, tool-call passthrough, MCP broker
custom provider chat, tool-call passthrough, MCP broker
chat/edit/apply, autocomplete, embeddings/context, Agent MCP
custom provider chat and /goal judge profile
Operators record live-client evidence from the dashboard or CLI. The public claim pack tells users which wording is allowed and which proof is still missing.