Documentation
Docs/en/reference/ai-control-protocol/index.md
Generated AiControl Protocol Reference
Generated reference. Do not edit directly. Update
BuildTools/AiControlProtocol.json, then runpython BuildTools/docs_ai_control_protocol.py --write.
| Index | Wire | Methods | Commands and events | Security | Integration and validation | Canonical JSON | Guide |
This reference defines the Engine-owned, project-neutral envelope for opt-in AI observation and control bridges. It does not define a game schema, MCP namespace, listener inside the core runtime, or server-authority bypass.
Contract status
| Field | Value |
|---|---|
| Stability | experimental |
| Support policy | The envelope is versioned but experimental; embedding projects must pin an Engine revision and own their observation and action schemas. |
| Protocol version | 1 |
| JSON-RPC marker | 2.0 |
| Default endpoint | 127.0.0.1:43011 |
| Maximum JSON payload | 1048576 |
| Stable entries | 49 |
| Source manifest | BuildTools/AiControlProtocol.json |
| Contract digest | d23079d2dda2357f9293250a395817f576437531c970245e710c436880bf8c65 |
| Reference | Purpose |
|---|---|
| Wire | Framing, envelopes, ordering, and error codes. |
| Methods | The six transport methods and their results. |
| Commands and events | Common command fields and asynchronous completion. |
| Security | Loopback, tokens, shipping builds, and authority. |
| Integration and validation | Project ownership and executable evidence. |
Boundary
Included:
- UTF-8 newline-delimited JSON over a TCP byte stream
- JSON-RPC-shaped request, result, and error envelopes
- authorization, liveness, status, observation, event, and action methods
- bounded transport, command queue, and event history behavior
- accepted-command and asynchronous completion lifecycle
- loopback-first threat boundary and reference validation
Excluded:
- a listener compiled into the FOnline core runtime
- project observation fields, game action names, entity semantics, and readiness gates
- MCP tool names, orchestration recipes, game-playing policies, and model prompts
- server authority bypasses, administrator commands, TLS, discovery, and internet exposure