You built the agent. But users are in Slack. Another team wants Teams. Legal needs audit logs. Product wants to try a different model. Each requirement means more plumbing you didn't sign up for.
Chat Agent Relay is that plumbing, done once. Connect any chat platform to any agent — your internal agent, a LangChain pipeline, a self-hosted model, or a commercial API. Governance, audit trail, zero lock-in.
Webhook parsing. Message formatting. Retry logic. Error handling. Audit logging. You write it for Slack. Then you write it again for Teams. Then again when you switch agents. This code is not your product. It's overhead — and it multiplies every time requirements change.
Every message flows through a 7-event chain. Chat platforms and agents are pluggable adapters on both sides. Swap either without touching business logic.
Chat Platforms Chat Agent Relay Your Agents ┌──────────┐ ┌──────────────────────────┐ ┌────────────────┐ │ Slack │ │ Canonical Event Chain │ │ Internal agent │ │ Teams │───>│ Policy + Governance │───>│ LangChain agent │ │ Discord │ │ Routing + Dispatch │ │ Self-hosted LLM │ │ Web │ │ Audit Ledger │ │ Any HTTP agent │ └──────────┘ └──────────────────────────┘ └────────────────┘
Slack and WebChat built-in. Add Teams, Discord, Telegram, or your custom surface with one interface.
Your internal agent, a LangChain pipeline, a self-hosted model, or a commercial API. Configure custom headers, request format, and response mapping — no need to modify your agent.
Keyword/regex policy engine blocks messages before they reach your agent. Every decision logged.
Every message is a 7-event chain in an append-only ledger. Query conversations and audit explanations via REST API.
SSE streams update chat messages progressively. Users see the agent thinking in real-time.
Exponential backoff on failures. Dead-letter queue for exhausted retries. No lost messages.
You built the agent. CAR handles delivery to Slack, Teams, web — with governance and audit you'd otherwise build from scratch.
Multiple teams, multiple agents, multiple chat surfaces. One middleware layer with consistent policy and observability.
Every AI interaction is an immutable event in an append-only ledger. Auditable by default. No "we'll add logging later."
You can always build the plumbing yourself. Here's what that means in practice.
| Capability | DIY Integration | CAR |
|---|---|---|
| New chat platform | Weeks of integration work | Implement one interface |
| Switch agents | Rewrite integration logic | Change one config |
| Audit trail | Build from scratch | Built-in event ledger + REST API |
| Policy / governance | Custom middleware | Configurable policy engine |
| Delivery retry | Implement yourself | Exponential backoff + DLQ |
| Streaming | SSE parsing + update logic | Progressive message updates |
| Time to production | Weeks | Minutes |
We're expanding both sides of the relay. Tell us which integrations matter to you.
Microsoft Teams, Discord, Telegram, WhatsApp Business, LINE — and a web widget SDK for embedding.
Any SSE-streaming agent, webhook/async agents, MCP-compatible agents, multi-agent fan-out routing.
RBAC, multi-tenant routing, usage analytics, admin UI for policy and routing management.
Which chat platform do your users live in? What kind of agent do you want to connect?
Tell us →