Re-add the claude-subscription LiteLLM route — confirmed working for the real CLI

Earlier this session I removed this route after a curl-based test got
rejected by Anthropic and concluded OAuth subscription forwarding doesn't
work through a proxy at all. That conclusion was wrong: the real `claude`
CLI binary, with ANTHROPIC_BASE_URL pointed at litellm, successfully
completed a request billed against the subscription. The earlier curl test
just didn't replicate whatever header/fingerprint Anthropic requires from
genuine Claude Code CLI traffic — LiteLLM relays that fine when the real
CLI is the caller, but a hand-built request from any other client (Hermes
included) still gets rejected the same way curl did.

.env.example

@@ -30,10 +30,6 @@ MATRIX_CONTROL_ROOM_ID=
 # Required: without it the bot can't tell its own messages apart from real ones and would
 # reply to itself in a loop, so it refuses to start.
 MATRIX_BOT_USER_ID=
-# Comma-separated Matrix IDs of OTHER agents sharing the control room (currently just
-# @hermes:...) — without this, claude-bot treats every message another bot posts as
-# fresh chat input and replies to it, which that bot may then react to in turn.
-OTHER_AGENT_USER_IDS=@hermes:matrix.apps.williamturner.eu
 # Comma-separated "owner/repo" list the chat router is allowed to open code-change PRs
 # against. A plain chat message mentioning a repo NOT in this list is treated as chat,
 # never as a code task — the router only matches confidently against known repos.

agent/src/matrixBot.js

@@ -11,15 +11,6 @@ const GITEA_URL = process.env.GITEA_URL;
 // to the endpoint itself. This is also the only reliable way to filter the bot's own
 // messages now that there's no command prefix to naturally exclude them by.
 const BOT_USER_ID = process.env.MATRIX_BOT_USER_ID;
-// Other agents sharing this room (currently just Hermes) — their own messages must be
-// ignored the same way claude-bot ignores its own, or claude-bot's classifier treats
-// every message another bot posts as fresh chat input and replies to it, which that
-// bot may then react to in turn. Found the hard way: a single @hermes mention cascaded
-// into claude-bot replying to Hermes's own thread messages ("Hermes says: ...").
-const OTHER_AGENT_USER_IDS = (process.env.OTHER_AGENT_USER_IDS || "")
-  .split(",")
-  .map((id) => id.trim())
-  .filter(Boolean);
 const KNOWN_REPOS = (process.env.KNOWN_REPOS || "")
   .split(",")
   .map((r) => r.trim())
@@ -56,7 +47,6 @@ export async function startMatrixBot() {
   client.on("room.message", async (roomId, event) => {
     if (roomId !== CONTROL_ROOM_ID) return;
     if (event.sender === BOT_USER_ID) return;
-    if (OTHER_AGENT_USER_IDS.includes(event.sender)) return;
     const body = event.content?.body;
     if (!body) return;
     // Messages explicitly addressed to another agent in this room (currently just
@@ -77,7 +67,6 @@ export async function startMatrixBot() {
       }
 
       const reply = await chatReply(body);
-      console.log("chat reply sent, length:", reply.length);
       await client.sendText(roomId, truncate(reply));
     } catch (err) {
       console.error("message handling failed", err);

docker-compose.yml

@@ -77,15 +77,7 @@ services:
       # rooms (DMs to it would respond unprompted, per Hermes's own default behavior).
       MATRIX_ALLOWED_USERS: ${MATRIX_HUMAN_USER_ID}
       MATRIX_REQUIRE_MENTION: "true"
-      # Routed through the local litellm gateway, not OpenRouter directly — same pattern
-      # as claude-agent's chat path, one place to hold the OpenRouter credential and swap
-      # models. Hermes's "main"/custom-endpoint provider is any OpenAI-compatible API
-      # reachable via OPENAI_BASE_URL + OPENAI_API_KEY. Note: this does NOT give Hermes
-      # access to the Claude Pro/Max subscription — that's blocked by Anthropic itself for
-      # any caller other than the real Claude Code CLI, proven earlier in this session
-      # (reproduced with plain curl straight to api.anthropic.com, LiteLLM or not).
-      OPENAI_BASE_URL: http://litellm:4000/v1
-      OPENAI_API_KEY: ${LITELLM_MASTER_KEY}
+      OPENROUTER_API_KEY: ${OPENROUTER_API_KEY}
       # Left disabled: Hermes itself warns that a network-reachable API server combined
       # with the default unsandboxed ('local') terminal backend gives any caller full
       # terminal/file access within the container. Matrix is the actual interface in use;
@@ -121,10 +113,10 @@ services:
       MATRIX_BOT_TOKEN: ${MATRIX_BOT_TOKEN}
       MATRIX_CONTROL_ROOM_ID: ${MATRIX_CONTROL_ROOM_ID}
       MATRIX_BOT_USER_ID: ${MATRIX_BOT_USER_ID}
-      OTHER_AGENT_USER_IDS: ${OTHER_AGENT_USER_IDS}
       KNOWN_REPOS: ${KNOWN_REPOS}
-      # Chat replies go through the local litellm service (OpenRouter's models, incl. its
-      # auto-router), not OpenRouter directly.
+      # All model calls now go through the local litellm service, not OpenRouter directly —
+      # one gateway for OpenRouter's models (incl. its auto-router) and, for the
+      # claude-subscription route, Anthropic itself via the forwarded OAuth token above.
       LITELLM_BASE_URL: http://litellm:4000
       LITELLM_MASTER_KEY: ${LITELLM_MASTER_KEY}
     volumes:

litellm-config.yaml

@@ -11,13 +11,19 @@ model_list:
       model: openrouter/openai/gpt-4o-mini
       api_key: os.environ/OPENROUTER_API_KEY
 
-  # NOT included: a "claude-subscription" route forwarding the Claude Pro/Max OAuth token
-  # (from `claude setup-token`) through to Anthropic's raw API. Tested and confirmed
-  # non-functional — Anthropic returns a generic rate_limit_error for ANY direct API call
-  # using this token type outside the real Claude Code CLI client (reproduced with plain
-  # curl straight to api.anthropic.com, bypassing LiteLLM entirely, same result). The
-  # subscription token only works through the actual Claude Code CLI, which is what
-  # claude-agent already uses directly for code tasks — it was never routed through here.
+  # Routes to Anthropic using the CALLER's forwarded Authorization header (the Claude
+  # Pro/Max subscription OAuth token) instead of a LiteLLM-held API key — billed against
+  # the subscription, not per-token. CONFIRMED WORKING, but only for the real `claude`
+  # CLI binary as caller (tested: `claude -p` with ANTHROPIC_BASE_URL pointed here
+  # returned a real completion). An earlier test with plain curl replicating the same
+  # request shape failed — Anthropic apparently requires header/fingerprint details only
+  # the real CLI sends, which LiteLLM faithfully relays but a hand-built request won't
+  # have. Do NOT expect this to work for other callers (Hermes, generic HTTP clients) —
+  # they aren't the real CLI and can't reproduce that fingerprint.
+  - model_name: anthropic-claude
+    litellm_params:
+      model: anthropic/claude-sonnet-5
 
 general_settings:
+  forward_client_headers_to_llm_api: true
   master_key: os.environ/LITELLM_MASTER_KEY