Unable to use Telegram in nemoclaw sandbox despite configured policy (engine:ssrf denial)

[huichia]

I have configured the Telegram policy in nemoclaw, but I still cannot use Telegram for conversations.

In the openshell logs, I see the following error:

[1776268872.421] [sandbox] [OCSF ] [ocsf] NET:OPEN [MED] DENIED /usr/local/bin/node(179) -> api.telegram.org:443 [policy:- engine:ssrf]

Environment

• nemoclaw v0.0.15
• openshell 0.0.25

Sandbox Configuration

telegram:
  name: telegram
  endpoints:
    - host: api.telegram.org
      port: 443
      protocol: rest
      enforcement: enforce
      tls: terminate
      rules:
        - allow: { method: GET, path: "/bot*/**" }
        - allow: { method: POST, path: "/bot*/**" }
        - allow: { method: GET, path: "/file/bot*/**" }

binaries:
  - { path: /usr/local/bin/node }
  - { path: /usr/bin/curl }

Question:
What additional configuration is required to make Telegram work properly in the sandbox?
Even though the policy already allows api.telegram.org, the connection is still blocked by engine:ssrf, causing the request to fail.

[nv-cyuen]

Hi huichia,

For a sanity check, did you configure your Telegram bot token and User ID during the onboarding process?

[huichia]

Hi @nv-cyuen ,

Yes, I've completed the onboarding steps:

• Obtained the Telegram bot token via @Botfather
• Obtained my User ID via @userinfobot
• Both have been configured, and I can confirm the token is present in ~/.nemoclaw/credentials.json under TELEGRAM_BOT_TOKEN

I also noticed that telegram-bridge is missing entirely from my nemoclaw status output:

nvidia@thor-head:~$ nemoclaw status
  Sandboxes:
    test * (nvidia/nemotron-3-nano-30b-a3b)
  ● cloudflared  (stopped)

While other users on older versions show it listed (even if stopped):

  ● telegram-bridge  (stopped)
  ● cloudflared  (stopped)

Could the missing telegram-bridge be related to the engine:ssrf denial? Is this expected behavior in the current version?

[nv-cyuen]

Hi @huichia,

Can you try:

openshell provider get <sandbox-name>-telegram-bridge

on the latest version of NemoClaw, and send me what you see?

[Uruglee2]

Hello et al,

I am also trying to get Telegram to talk to openclaw via Nemoclaw. All I have got for the past two days is a cloud inference working. I want to run this local but I have this persistent error in the sandbox that just will not let me connect to telegram. I am using a Jetson ORIN AGX 64Gb Dev Kit with 1tb ssd.

The Sandbox output:

🦞 OpenClaw 2026.4.9 (0512059) — I can grep it, git blame it, and gently roast it—pick your coping mechanism.

│
◇
07:24:12+00:00 [gateway] loading configuration…
07:24:12+00:00 [gateway] resolving authentication…
07:24:12+00:00 [gateway] starting...
07:24:13+00:00 [gateway] failed to persist plugin auto-enable changes: Error: EACCES: permission denied, open '/sandbox/.openclaw/openclaw.json.4604.481bd78a-c522-4767-bb32-9f6d43aa6998.tmp'
07:24:13+00:00 [gateway] starting HTTP server...
07:24:13+00:00 [canvas] host mounted at http://127.0.0.1:18790/__openclaw__/canvas/ (root /sandbox/.openclaw/canvas)
07:24:13+00:00 [gateway] MCP loopback server listening on http://127.0.0.1:40529/mcp
07:24:13+00:00 [tasks] Failed to restore task registry
07:24:13+00:00 [heartbeat] started
07:24:13+00:00 [health-monitor] started (interval: 300s, startup-grace: 60s, channel-connect-grace: 120s)
07:24:13+00:00 [gateway] agent model: ollama/qwen2.5-coder:32b
07:24:13+00:00 [gateway] ready (6 plugins, 1.3s)
07:24:13+00:00 [gateway] log file: /tmp/openclaw-998/openclaw-2026-04-29.log
07:24:13+00:00 [gateway] starting channels and sidecars...
07:24:13+00:00 [model-pricing] pricing bootstrap failed: TypeError: fetch failed
07:24:14+00:00 [hooks] loaded 4 internal hook handlers
07:24:23+00:00 [bonjour] watchdog detected non-announced service; attempting re-advertise (gateway fqdn=sealz-assistantz (OpenClaw)._openclaw-gw._tcp.local. host=openclaw.local. port=18790 state=probing)
07:24:24+00:00 [plugins] embedded acpx runtime backend registered (cwd: /sandbox/.openclaw/workspace)
07:24:24+00:00 [browser] control listening on http://127.0.0.1:18792/ (auth=token)
07:24:25+00:00 [telegram] [default] starting provider (@Uruglee2_Nemoclaw_bot)
07:24:45+00:00 [agent] embedded run agent end: runId=7d226f85-c939-4162-b940-60edb10ff3e3 isError=true model=qwen2.5-coder:32b provider=ollama error=LLM request failed: network connection error. rawError=fetch failed | Request was cancelled. | Proxy response (403) !== 200 when HTTP Tunneling
07:24:47+00:00 [agent] embedded run agent end: runId=7d226f85-c939-4162-b940-60edb10ff3e3 isError=true model=qwen2.5-coder:32b provider=ollama error=LLM request failed: network connection error. rawError=fetch failed | Request was cancelled. | Proxy response (403) !== 200 when HTTP Tunneling
07:24:51+00:00 [agent] embedded run agent end: runId=7d226f85-c939-4162-b940-60edb10ff3e3 isError=true model=qwen2.5-coder:32b provider=ollama error=LLM request failed: network connection error. rawError=fetch failed | Request was cancelled. | Proxy response (403) !== 200 when HTTP Tunneling
07:24:59+00:00 [agent] embedded run agent end: runId=7d226f85-c939-4162-b940-60edb10ff3e3 isError=true model=qwen2.5-coder:32b provider=ollama error=LLM request failed: network connection error. rawError=fetch failed | Request was cancelled. | Proxy response (403) !== 200 when HTTP Tunneling
07:24:59+00:00 [agent] embedded run failover decision: runId=7d226f85-c939-4162-b940-60edb10ff3e3 stage=assistant decision=surface_error reason=auth provider=ollama/qwen2.5-coder:32b profile=-
07:25:00+00:00 [telegram] sendMessage ok chat=8644173652 message=287

At the least I need someone to point me in the right direction as I feel I am getting to far away from the real issue and can not see it anymore.

Kind regards

L

[nv-cyuen]

Hi @Uruglee2,

Can you run

curl -v https://inference.local/v1/models

inside your sandbox just to see if it can connect properly and let me know how that goes first?