Select a topic
Messages stay on this serverT/
Start a conversation
Create a topic to keep each thread of work in its own place.
Create a topic to keep each thread of work in its own place.
Change the password used to open this inbox. Other signed-in browsers will be signed out.
Run these commands on the machine with Hermes installed, not on the TeleIn server. Only outbound HTTPS is needed; no SSH access to TeleIn is required.
Install a current Hermes release on that machine first. Create a new profile, or skip creation if your dedicated profile already exists. Do not use the default profile or clone messaging credentials.
Keep this profile separate from the default multiplexer:
Choose your provider and model, and sign in or enter its API key in Hermes. These credentials stay on the Hermes machine. A new profile does not inherit the default profile's model credentials.
For a custom OpenAI-compatible endpoint that supports /v1/responses, choose Responses API. This is required when reasoning and function tools are used together on endpoints such as Amazon Bedrock's OpenAI-compatible API.
Run a model-only check. Continue only after it returns a real reply, without a 401 or authentication error.
The TeleIn bot token is not a model API key. Never paste provider credentials into TeleIn.
These settings select this server's Bot API and allow only its owner to message Hermes. They preserve pending messages on startup.
Locate this profile's secrets file, then open the printed path in your text editor:
Paste the copied line into that file, replacing any existing TELEGRAM_BOT_TOKEN line, then save. Keep the file private and clear your clipboard afterward. Do not paste this line into a terminal, chat, or issue report.
Allow this private profile to use Hermes tools without pausing for routine command approval:
This applies only to the selected profile. Hermes still enforces its non-bypassable hardline protections.
Keep this terminal open and wait for Telegram polling to connect. For unattended use, configure a service for this profile on your Hermes machine.
Use one gateway per TeleIn bot token. When moving machines, stop only the old TeleIn profile gateway first. Do not force startup, replace another gateway, or restart the default gateway.
Wait for the next profile rescan (usually within 30 seconds). If it remains blocked, open Hermes's managed Python console:
hermes --run-module codeAt the Python prompt, paste the following. Confirm the returned list excludes your TeleIn profile before retrying startup. This does not restart other gateways.
This sends a marked message in General and waits up to two minutes for the expected reply. It uses your configured model and may incur provider usage charges.
No reply check has been run.
An update leaving the queue confirms neither model processing nor reply delivery. This check passes only when TeleIn stores the expected reply. TeleIn cannot inspect credentials on your Hermes machine.
/v1 root; Hermes adds /responses.