Crate link_assistant_router 

Link.Assistant.Router — Claude MAX OAuth proxy and token gateway.

A Rust-based API gateway that proxies Anthropic (Claude) APIs, supports Claude MAX OAuth sessions, and provides multi-tenant access via custom-issued tokens.

Modules

accounts

Multi-account OAuth credential routing.

accounts_cli

The accounts subcommand: report what each configured account can do.

activitypub

Minimal ActivityPub / ForgeFed endpoint support.

admin

Admin credential claim state — the two-phase first-visitor bootstrap.

admin_api

HTTP surface of the admin UI: the bootstrap claim, credential rotation, and read-only status, plus the router that serves them on the dedicated admin port.

admin_auth

Authorisation decision for the administrative endpoints.

admin_config

Configuration for the opt-in admin UI listener.

admin_ui

Serving of the embedded admin UI bundle.

anthropic_bridge

Anthropic Messages surface over non-Anthropic upstreams.

anthropic_stream

Incremental OpenAI SSE to Anthropic SSE translation.

app_state

Shared state used by HTTP route handlers.

audit

Append-only per-token audit log.

auth

Provider-aware native Claude and Codex subscription authorization.

auth_remote

Authorize the subscription of the selected router, not a local directory.

bridge_selection

Deterministic bridge-model selection from live provider catalogs.

capabilities

Central capability declarations for request reconciliation and bridges.

chat_admin

Transport-agnostic core of the optional chat admin channels.

chat_commands

The administrative commands reachable from a chat channel.

chat_config

Configuration for the optional chat admin channels.

claude_auth

Native Claude OAuth authorization using the public Claude Code client.

claude_identity

Claude Code request identity used for OAuth inference and account probes.

cli

Command-line interface for the router.

client_command

Output and dispatch layer for the clients CLI command.

client_global

Exact, edit-aware backup and undo for router with --global.

client_policy

Client-bound authorization for consumer subscription credentials.

clients

Safe local configuration for agentic CLI clients.

codex_identity

Exact HTTP identity used by the supported Codex client.

codex_loopback_bridge

Loopback transport accepted by Codex’s remote-control origin policy.

config

Configuration module for Link.Assistant.Router.

config_defaults

Provider-specific default constructors used by crate::config.

configure

Permanent client setup: one name, one targeting rule, one reversal.

conversation_record

Opt-in recording of client-to-Router conversations, and strict replay of one.

conversations

Authenticated native OpenAI Conversations and conversation-items lifecycle.

crater

Crater ForgeFed task provider.

credential_acceptance

Staged, non-inference acceptance for credentials that may replace a primary.

credential_recovery_store

Durable fallback storage for refreshed subscription credentials.

credential_status

Shared local/remote subscription credential acceptance checks.

credential_store

Where a subscription credential lives, so every refresh path can re-read it, lock it, and write a rotated refresh token back.

doctor

Subscription diagnostics shared by the doctor CLI command.

durable_file

Owner-only, crash-durable file replacement and inter-process locking.

entrypoint

Running the asynchronous entry point on a stack this crate controls.

env_paths

Reading directory paths from the environment.

gemini

Google Gemini (Code Assist) subscription upstream.

gemini_bridge

Native Gemini ↔ OpenAI Chat Completions translation.

git_proxy

Git smart-HTTP proxy with ref-update policy.

github_proxy

GitHub API credential proxy with a deny-by-default destructive policy.

gonka

Gonka upstream provider support.

lefine

Lefine’s persisted OpenAI Chat Completions provider contract.

lino_json

Persist serde-shaped state as readable links notation.

log_analysis

Answer questions about the request log.

log_decode

Decoding of stored response bodies for log inspection.

logging

Tracing filter construction shared by the binary and regression tests.

login

Login sessions that authorize a deployment without a pre-existing credential file.

login_api

HTTP endpoints for authorizing a deployment (issue #47).

login_pty

Long-lived PTY sessions for driving an interactive terminal program.

login_url

Extraction of the authorization URL and the resulting token from a terminal transcript.

managed_server

Server selection, managed Docker lifecycle, and per-run token handling.

metrics

Live metrics and operational endpoints.

model_catalog

Live subscription model discovery with stale-on-error caching.

model_resource

Authenticated single-model retrieval from native or fresh live catalogs.

model_routing

Model catalog and automatic subscription-provider routing.

monitoring_api

Monitoring endpoints served on the proxy port.

mpp

Machine Payments Protocol helpers for paid OpenAI-compatible endpoints.

oauth

Claude MAX OAuth credential reader.

on_demand_cli

Disposable package-runner environments for vendor CLI compatibility.

openai

OpenAI-compatible API surface.

output_limit

Local emulation of max_output_tokens for backends that reject the field.

platform_keychain

The platform secret store a vendor CLI keeps its live subscription in.

primary_listener

Explicit primary listener configuration.

provider_acceptance

Transactional acceptance for persisted upstream providers.

provider_proxy

OpenAI-compatible provider API and forwarding helpers.

providers

Provider configuration storage.

providers_cli

router providers — manage stored OpenAI-compatible providers.

proxy

Transparent API proxy for forwarding requests to upstream APIs.

refresh

OAuth refresh and durable rotation handling for vendor subscriptions.

refresh_rejections

Which credentials a refresh has already been refused for, across restarts.

remote_command

One targeting rule for every command that reads or changes router state.

request_log

Redacted, size-bounded HTTP exchange logging.

responses

OpenAI Responses-API (POST /v1/responses) request/response translation.

responses_lifecycle

Stored OpenAI Responses resource lifecycle proxying.

responses_websocket

Native OpenAI Responses WebSocket forwarding with a pinned provider.

route_contract

Canonical HTTP route classification and endpoint construction.

security_headers

Response hardening for the admin listener.

server_command

CLI output layer for router server.

server_router

Network-facing route construction from the canonical route contract.

stop_sequences

Local stop-sequence enforcement for backends that cannot accept stop.

storage

Token persistence backends.

subscription

Subscription OAuth credential readers for vendor coding CLIs.

subscription_health

Subscription health reporting (issue #318).

subscription_proxy

Forward OpenAI-style requests to vendor subscription upstreams.

subscription_usage

subscription_usage_cli

Formatter and HTTP client shared by local and selected-remote usage commands.

telegram

Telegram admin channel — a long-polling loop over the Bot API.

tls

Optional TLS termination in the router itself.

tls_cli

router tls — manage the certificate a self-signed deployment serves.

token

Custom token management for the gateway layer.

token_admin

Admin HTTP endpoints for managing router-issued tokens.

token_report

How router tokens list renders, for local and remote alike.

token_reservation

Pre-dispatch spend reservations for per-token max_tokens budgets.

token_secret

The stand-in used when a command will never sign anything.

tokens_remote

router tokens against the selected router.

unix_listener

Serving the router on a unix domain socket.

upstream_client

Construction of the shared HTTP client used for every upstream call.

usage

Upstream usage extraction for per-token spend budgets.

vendor_cli_refresh

Last rung of the recovery ladder: let the vendor’s own CLI rotate the chain.

vk

VK admin channel — a long-polling loop over the Bots Long Poll API.

with_command

Temporary-by-default launcher shared by router with and with-router.

zai_coding_plan

Policy and exact model registry for the experimental z.ai GLM Coding Plan.

Constants

VERSION

Package version (matches Cargo.toml version).