SELF-HOSTED MESSAGING · BUILT FOR CONTROL

Deploy messaging and AI supporton your own infrastructure

One Go binary powers group chats, live rooms and DMs. Your data, keys and access policies stay under your control — with a free edition to get started.

Try the live demoSee the quickstart
  • Never priced per DAU
  • Intranet and offline ready
  • Built to be customized
Product preview: the AI agent calls a business tool, then hands the full context to a human agent.

Real benchmarks · reproducible by you

10,000
concurrent connections
38,506
msg/s fan-out
153 ms
p99 latency
0
errors / drops

ONE ENGINE, THREE CONVERSATION MODES

More than a chat box — a messaging system you control

User-to-user messaging, high-concurrency live interaction, and AI-to-human support share one identity, message, and permission model.

GROUPS & DMs

Make conversation native to your product

Groups, DMs, replies, recall, read receipts and reactions work out of the box across Web, iOS, Android and server SDKs.

  • Complete messaging
  • Multi-platform SDKs
  • Offline push

LIVE ROOMS

High concurrency without losing control

Room messages, presence, moderation and rate limits run through one reliable path.

  • Presence
  • Moderation
  • Rate limits

AI CUSTOMER SERVICE

AI resolves first. Humans step in naturally.

RAG, tool calls, memory and human handover live directly on the message stream — no second support stack to stitch together.

  • RAG
  • Tool calls
  • Human handover
  • Audit log
  1. 01Understand
  2. 02Call business tools
  3. 03Answer or hand over

FROM ZERO TO PRODUCTION

Three steps to own your messaging stack

Validate with one process, then scale with real traffic. Development and production keep the same protocol and code.

  1. 01

    Boot

    Connect one Go binary to PostgreSQL and Redis and get running in minutes.

  2. 02

    Integrate

    Clients open one WebSocket; your backend syncs identities and events over gRPC.

  3. 03

    Scale

    Add RabbitMQ or NATS and scale horizontally without rewriting clients; use Kubernetes + Helm for production rollouts and autoscaling.

Run the quickstart yourself
Up in one minute
imcore — zsh
# build and start
$ make build-basic
$ ./bin/basic/websocket

✓ http+ws  :8080
✓ grpc     :9001
✓ metrics  :9090

● IMCore is ready

Real benchmarks · no inflated claims

Single-node performance: reproducible numbers, not theoretical peaks

Measured on Alibaba Cloud cn-hangzhou, 2026-07. Different instance types, message sizes and room topologies will vary — rerun the benchmark for your own workload shape.

10,000
concurrent connectionsrecommended low-latency step, zero errors / drops
38,506
msg/s fan-outactual group-message delivery rate
153 ms
p99 latencyp50 just 83ms
0
errors / dropsno slow-consumer closes, no throttling
Test environment (single instance)
Instance
Alibaba Cloud ECS · 4 vCPU / 8 GiB (hfc6.xlarge)
Process
One websocket binary · single-node direct delivery (no broker)
Storage
PostgreSQL + Redis (same host)
Scenario
Multi-room group chat (50 rooms) · steady step of a saturation sweep
Tuning
reliable_send_buffer=256 (default) · GOGC=200

Why these numbers hold up

  • We hand you the load-test tooling

    We provide the load generator and report tool (loadtest / satreport) so you can rerun the exact parameters, scenarios and result JSON in your own environment — reproducible on your side, not just our word.

  • Modelled on real load

    Traffic is shaped like real users typing (within the server’s 4 msg/s per-user limit), not one connection hammering to inflate throughput.

  • Steady values, not the edge of collapse

    We publish the zero-error, zero-drop low-latency step — 10,000 conns, p99 153ms, zero drops (consistent across two independent reruns). Going higher means scaling out horizontally (next point), not squeezing one node to the edge.

  • A single node is just the floor

    RabbitMQ / NATS scale out horizontally; capacity grows near-linearly with websocket instances — multiply the single-node figure by your instance count.

ONE CODEBASE · SINGLE NODE TO CLUSTER

Scale the business without rebuilding the stack

The client protocol stays the same. Start with direct delivery, then add RabbitMQ or NATS to scale websocket, persistence, AI and sync workers.

SECURITY IS THE DEPLOYMENT MODEL

Security & private deployment

Instead of trusting another hosted vendor, run IMCore inside your own network, key-management and audit boundary.

Security & private deployment →

Data sovereignty · self-hosted, never leaves

Lifecycle controls · retention / archive / export / erasure

Audit protection · legal hold / HMAC chain / redaction

Recovery evidence · backup restore / RPO-RTO drills

START FREE · UPGRADE WITH REAL NEEDS

Start running, then upgrade for real needs

Basic stays free. Add clustering, the admin console or AI support only when you need them — never stack costs by DAU.

01

Basic

Single-instance IM core, self-hosted starter

Free forever
02

Pro

Cluster-scalable with push / moderation / callbacks — production-ready

¥9,800/ year
03Recommended

Enterprise

Pro + admin console + Community (super-groups & channels) + external sync

¥29,800/ year
04

Customer Service

Pro + the full AI customer-service suite

¥19,800/ year
See full feature comparison and pricing

READY TO OWN THE CONVERSATION?

Bring your workload. We’ll size the architecture.

Share your concurrent connections, daily message volume and deployment environment for a concrete edition and topology recommendation.

We usually reply within one business day