How Agent Reach's Multi-Backend Routing Keeps Your AI Agent's Internet Access Working

Jun 16, 2026

Problem

Platform access tools break frequently. APIs change, anti-scraping measures evolve, CLI tools get abandoned. When I used yt-dlp to read Bilibili video subtitles, it worked for months — until Bilibili’s risk control system 412-blocked it.

I didn’t want to manually track which tool works for which platform. I wanted the system to handle it.

How Multi-Backend Routing Works

Agent Reach solves this with ordered backend lists. Each platform channel defines multiple backends in priority order:

class Channel(ABC):
    name: str = ""
    backends: List[str] = []
    active_backend: Optional[str] = None

    def ordered_backends(self, config=None) -> List[str]:
        candidates = list(self.backends)
        override = config.get(f"{self.name}_backend") if config else None
        if override:
            for i, b in enumerate(candidates):
                if b == override or b.startswith(override):
                    candidates.insert(0, candidates.pop(i))
                    break
        return candidates

The check() method doesn’t just check if a binary exists via shutil.which(). It really probes each backend by running commands and parsing output. The first backend returning “ok” becomes active.

Real-World Example: Bilibili

Here’s the Bilibili channel’s backend configuration:

backends = ["bili-cli", "OpenCLI", "B站搜索 API"]

When Bilibili 412-blocked yt-dlp in June 2026, yt-dlp was removed from the backend list entirely (documented in a comment at the top of the file). The new primary bili-cli was already in position. Users experienced zero disruption — the system had already switched.

Bilibili backends (June 2026):
┌──────────┐   ┌──────────┐   ┌──────────┐
│ bili-cli │ → │ OpenCLI  │ → │Search API│
│ (active) │   │ (fallback)│   │(last resort)
└──────────┘   └──────────┘   └──────────┘

Twitter’s Backend Chain

backends = ["twitter-cli", "OpenCLI", "bird CLI (legacy)"]

Twitter’s probe has a two-stage check: it collects all candidates first, then prefers “ok” over “warn”. This prevents an installed-but-not-logged-in twitter-cli from blocking a working OpenCLI.

Environment-Aware Routing: XiaoHongShu

backends = ["OpenCLI", "xiaohongshu-mcp", "xhs-cli (xiaohongshu-cli)"]

On desktop, OpenCLI reuses your Chrome browser session — zero friction. On a server, OpenCLI naturally fails the probe (no desktop Chrome), so xiaohongshu-mcp (headless browser) takes over. The routing is environment-aware without hardcoding environment checks.

Backend Override

If you need to force a specific backend:

agent-reach configure twitter_backend "OpenCLI"

agent-reach doctor --json

Unknown override values are ignored. This prevents stale configs from hiding working backends.

Summary

In this post, I explained how Agent Reach’s multi-backend routing keeps your AI agent’s internet access working even when individual tools break. The key point is that each platform has an ordered list of backends, and check() probes each one by real command execution. Platform changes are handled at the routing layer — no reconfiguration needed, no disruption, zero user action.

Final Words + More Resources

My intention with this article was to help others share my knowledge and experience. If you want to contact me, you can contact by email: Email me

Here are also the most important links from this article along with some further resources that will help you in this scope:

Oh, and if you found these resources useful, don’t forget to support me by starring the repo on GitHub!