Chats & Conversations
One inbox for every conversation. Chats unify SMS, voice calls, web chat, and email into a single timeline—giving your team complete context for every member interaction.
What Chats Do
Chats are your unified communication hub:
- Single timeline for all channels (SMS, voice, web, email)
- Full history with messages, transcripts, and events
- AI + human collaboration in the same thread
- Smart routing based on labels and rules
- Real-time updates as conversations happen
Key Features
Multi-Channel Threading
Unlike siloed systems, Gravity Rail keeps all interactions together:
| Channel | Same Chat |
|---|---|
| Web chat | ✓ |
| SMS | ✓ |
| Voice calls (transcribed) | ✓ |
| ✓ |
Human Takeover
Seamlessly transition from AI to human:
- AI handles initial conversation
- Staff joins when needed
- Human has full AI conversation context
- Member experiences smooth handoff
Smart Organization
- Labels tag chats (Urgent, VIP, Follow-up)
- Filters create saved views
- Search finds any conversation
- Assignment routes to the right person
AI Collaboration
AI assistants work alongside your team:
- Respond automatically when enabled
- Collect data via conversation
- Escalate on configured triggers
- Suggest responses for human approval
How It Works
Any Channel → Unified Chat → AI or Human → Complete History
(SMS/voice/web) (one thread) (responds) (searchable)
A member starts on web chat, continues via SMS, and calls in—all captured in one conversation. Your team sees everything in context.
Common Use Cases
| Use Case | What It Enables |
|---|---|
| Multi-Channel Support | Complete context across channels |
| Appointment Management | Reminders, confirmations, rescheduling |
| Sales Follow-Up | AI nurtures, humans close |
| Team Inbox | Shared visibility, clear ownership |
Approvals in chat
Some things an agent proposes are worth a second pair of eyes — writing a file, changing a workflow, sending on your behalf. When one comes up, the chat stops and shows you what it wants to do, in plain language: "The Manager wants to write to /plans/referral-enrollment/README.md in your workspace."
You have two answers.
Approve lets that one action run. Always approves this kind of action from now on in this workspace, so you stop being asked about it. Use it once you have seen enough of a given action to trust it.
Reject stops the action, and asks you why. That reason goes back to the agent, so rejection is how you steer: "not that folder — put it under /plans/2026/" gets you a corrected proposal you can read and approve, rather than the same one again.
Rejecting with a reason is deliberately the only way to change what happens. Hand-editing the agent's arguments would mean approving a call nobody re-reasoned about — easy to get subtly wrong, and it skips the agent's own checks. Telling it what you wanted keeps the agent responsible for the result.
Details expands the exact call — the tool and every argument — if you want to audit precisely what was proposed before deciding.
The prompt fills the chat while it is waiting, and Show chat tucks it into a bar at the bottom so you can read the conversation behind it. It never disappears: something is paused waiting for your answer, so there is no way to dismiss it without answering.
Pricing & Limits
Feature availability and limits are determined by your subscription plan. See the pricing page for details.
Related Resources
- Agents - Configure conversation AI
- Phone Numbers - SMS and voice setup
- Workflows - Structure conversations
- Labels - Organize and route chats