🚀 OwpenGram Server

Your own private messaging server — self-hosted, protocol-compatible, fully yours.

The protocol stack is built on the published github.com/iamxvbaba/td module (v1.1.0), using a canonical Layer 228 schema with sparse tlprofile exact Layer 225-228 compatibility profiles.

If you are looking for a Telegram server, MTProto server, Telegram backend, Telegram clone server, or self-hosted Telegram-like chat server, this repository is the server-side implementation to study, run, and improve.

OwpenGram Server is an open-source, Telegram-compatible MTProto backend written in Go. Run it on your own network for a private, closed setup, or on a VPS to be reachable anywhere in the world. Your data, your keys, your rules — no cloud, no lock-in, no censorship.

🔗 Implements MTProto API layer 228.

OwpenGram Server is independent and unofficial. It is not affiliated with, endorsed by, or sponsored by Telegram or the official Telegram team.

• 🔒 Private & self-hosted — messages live on infrastructure you control.
• 🧩 Telegram-compatible — works with the OwpenGram Android and Desktop clients.
• 🌍 Reachable anywhere — host it globally, or keep it on your own network.
• 🛡️ Censorship-resistant — no central authority can shut you down.
• ⚙️ Single binary — one Go program prepares keys, runs migrations, serves MTProto, and dispatches updates and background workers.
• 🆓 Free & open source — Apache-2.0, audit and extend it freely.

• 💬 Private chats, groups, supergroups & channels
• 📞 Voice & group calls, live streams, SFU/TURN building blocks
• 🖼️ Media & files — photos, videos, documents, stickers, reactions
• 🪣 Media storage on local disk or an S3/MinIO-compatible object store
• 🤖 Bots and mini apps, with a minimal Bot API gateway
• 🏷️ Fragment-style NFT/collectible usernames and bot verification marks
• 📢 Admin-panel broadcasts — announce to every user, or a picked list
• 🔑 Self-hosted "Log in with Telegram" (OpenID Connect) and passkey sign-in
• 🌐 Message translation and AI-assisted compose
• 📇 Contacts, dialogs sync, chat folders, public link landing pages
• 🖥️ Admin API and web UI for operations, plus a TUI server panel to run it all

Requirements:

• Go 1.25+
• Docker (or Docker Desktop), for PostgreSQL and Redis
• OpenSSL, to export the server's RSA public key for the client's "Add Server" dialog

1. Clone the repository

git clone https://github.com/owpengram/owpengram-server.git
cd owpengram-server

2. Start the infrastructure (PostgreSQL + Redis)

docker compose -f deploy/docker-compose.yml up -d

3. Build and run the server

Windows (PowerShell):

go build -o bin/gramsrv.exe ./cmd/telesrv
.\bin\gramsrv.exe

Linux / macOS:

go build -o bin/gramsrv ./cmd/telesrv
./bin/gramsrv

On first start, the server creates data/server_rsa.pem, applies database migrations, seeds bundled language packs, prepares optional media resources, starts MTProto on 0.0.0.0:2398, and brings up the update/media/background workers in the same process.

Default local login code: 12345 — change it before any real use!

💡 Prefer a menu over the command line? Steps 2 and 3 above (Docker infrastructure, build, run) can be done through the bundled TUI server panel instead — see "🖥️ Server Panel" right below.

A cross-platform interactive TUI wraps the steps above — Docker naming migration, docker compose up, go build, and launching both owpengram-server and owpengram-admin-panel — behind a menu, so you don't re-run commands from scratch every time.

./owpengram-server.sh     # Linux/macOS

.\owpengram-server.bat    # Windows

Both launchers check prerequisites first (Go, Python 3, and the panel's own dependencies via tui-panel/requirements-panel.txt), then start the panel.

What it does:

• 🧙 First-run setup wizard — walks through the required .env values before the first start.
• ▶️ Start / Stop / Restart — launches owpengram-server and owpengram-admin-panel as detached background processes; closing the panel does not stop them, only "Stop" does. Reopening the panel later picks the same processes back up and reports live status.
• ⬆️ Update — git pull --ff-only, rebuilds both binaries, restarts them, and re-execs the panel itself so it also picks up any change to its own code — one menu action instead of a manual pull/build/restart sequence.
• 📜 Live log viewer — tail either binary's log, or both in a split view.
• ⚙️ .env editor — edit configuration from inside the panel, grouped by feature, without hand-editing the file.

.env.example is the complete configuration reference — every variable is documented there with an explanatory comment, grouped by feature. Copy it to .env and edit, or set the variables directly as environment variables. Most commonly used variables:

Optional OpenAI-compatible, Kimi/Moonshot, Gemini, and Anthropic AI provider variables, login email/SMTP settings, and Business AI settings are documented in .env.example.

By default the server just uses a fixed development code (TELESRV_DEV_AUTH_CODE, 12345) for everything and writes it as a durable 777000 message on the account. Three optional, independent features build on top of that:

1. External code delivery providers — send the real login/phone-change code out via SMS webhook or SMTP instead of only the fixed dev code. An existing account's durable 777000 message is always written first; the provider is an additional delivery channel, never a replacement.

2. Login email — an extra verification factor on top of the phone number (not a replacement for it), similar to Telegram's own login-email feature.

3. Email-as-identity sign-up — lets patched clients register and log in with just an email address, no phone number at all. The client encodes the email into a synthetic 888-prefixed number and drives the ordinary sendCode/signUp/signIn/changePhone flow unchanged; the server decodes that number back to the email and delivers the code over the same channel as login email above (TELESRV_EMAIL_CODE_DELIVERY_PROVIDER). The account's real, visible phone number is a random-looking short number from TELESRV_EMAIL_SIGNUP_PHONE_PREFIXES — cosmetic only, never used to route the code — so admins can make freshly signed-up accounts look locally flavored (e.g. 888,380,373) without any client update.

Uploaded media (photos, documents, stickers) can live on this machine's disk, or in an S3-compatible object store. deploy/docker-compose.yml bundles a self-hosted MinIO container, pre-wired to the defaults below, so s3 (the default) works out of the box with no extra setup — point the same variables at AWS S3 instead if you'd rather not self-host it.

Switching backends only affects new uploads — existing files stay wherever they were written and remain reachable as long as that backend's settings stay filled in. MinIO's own web console is reachable at http://localhost:9001 (docker compose -f deploy/docker-compose.yml up -d also starts it) using the same access key/secret as above.

Related toggles (defaults in .env.example's Advanced section): a low-space guard that rejects new uploads once storage nears full, and automatic cleanup of media no longer referenced by any message.

When deploying on a public server, open the following according to the features you enable.

Minimal (chat only)

With admin backend

Optional features

Internal/debug (do not expose publicly)

Make sure TELESRV_LISTEN=0.0.0.0:2398 is set, and TELESRV_ADVERTISE_IP points to your public IP so clients can connect.

The server can serve public landing pages for /<username>, profile avatars, /addstickers/<shortName>, /addemoji/<shortName>, and /addlist/<slug>.

TELESRV_PUBLIC_LINK_WEB_ADDR=127.0.0.1:2401
TELESRV_PUBLIC_BASE_URL=https://your-domain.example
TELESRV_PUBLIC_APP_SCHEME=yourapp
TELESRV_PUBLIC_WEB_BASE_URL=https://web.your-domain.example
TELESRV_PUBLIC_APP_NAME=YourApp

In production, keep TELESRV_PUBLIC_LINK_WEB_ADDR on loopback and reverse-proxy the public routes to it with HTTPS.

Use the OwpenGram clients, which have a built-in Add Server option on the server-selection screen at login — no source patching or custom build needed:

• 🤖 Android client
• 💻 Desktop client

A stock Telegram client will not connect, since it only trusts Telegram's own DC list and RSA keys.

• Telegram Desktop commit: 9caf32dffc90ddd9bb08ad5777b865f729fa167b
• Canonical TL layer: 228
• Exact compatibility profiles: Layer 225-228
• Local DC: 127.0.0.1:2398, DC id 2

1. Export your server's public key

After the server generates data/server_rsa.pem, export the matching public key as PEM:

openssl rsa -in data/server_rsa.pem -RSAPublicKey_out -out data/server_rsa.pub

2. Add the server in the client

On the login screen, open server selection → Add Server, and fill in:

• Host — your server's address (e.g. 192.168.1.50 or chat.example.com)
• Port — 2398 by default
• Main data center — the DC id from TELESRV_DC (2 by default)
• RSA Public Key — paste the full contents of data/server_rsa.pub (the -----BEGIN RSA PUBLIC KEY-----... PEM block) into the key field

Use separate client working directories so sessions do not share local tdata:

$tdesktop = "C:\path\to\tdesktop\out\Debug\Telegram.exe"
Start-Process $tdesktop -ArgumentList @("-workdir", "$PWD\.tdata-alice")
Start-Process $tdesktop -ArgumentList @("-workdir", "$PWD\.tdata-bob")

Log in with different phone numbers — the local login code is 12345 unless you changed TELESRV_DEV_AUTH_CODE. Recommended checks:

• Send private messages, stickers, media, replies, forwards, edits, deletes, and read receipts between two users.
• Keep one device online and restart another device to verify offline updates.getDifference recovery.
• Open the same account from multiple sessions and confirm current-session echoes are not duplicated while other online sessions receive updates.
• Check server logs for no new NOT_IMPLEMENTED, Unhandled RPC, bad_msg, panic, or internal errors.

cmd/telesrv/              server entrypoint
cmd/telesrv-admin/        admin backend and web UI
tui-panel/                interactive TUI server panel (setup, start/stop, update, logs, .env editor)
deploy/                   docker-compose (incl. MinIO), migrations, deploy helpers
data/                     bundled language packs and optional seed data
internal/mtprotoedge/     MTProto transport, auth key, session, ack/resend
internal/rpc/             TL router and client compatibility handlers
internal/app/             domain services
internal/domain/          protocol-independent domain models
internal/store/           memory/postgres/redis storage backends
internal/seed/            bundled seed catalog loaders
internal/sfu/             real-time SFU experiments
internal/turnsrv/         TURN/STUN building blocks

This server gets better fastest with real usage and focused fixes:

• Telegram Desktop and Android compatibility reports with reproducible steps.
• RPC traces for startup, sync, chat, media, calls, bots, or edge cases.
• Focused fixes for implemented paths instead of broad rewrites.
• Tests for online/offline updates, multi-device sessions, read state, media, and channel behavior.
• Performance work on hot paths such as fan-out, pagination, storage queries, media upload/download, and connection handling.

If a change affects visible client behavior, please include the client version/commit, the RPC path you tested, and whether server logs stayed free of new NOT_IMPLEMENTED, Unhandled RPC, bad_msg, panic, or internal errors.

• 📢 Channel: @owpengram
• 💬 Chat: Join the discussion

OwpenGram Server builds on the open-source gramsrv project.

Apache License 2.0

⭐ If OwpenGram is useful to you, a star on GitHub helps the project grow.