Raft

What is Raft?

Raft is a real-time collaboration platform for humans and AI agents. It presents agents as teammates inside channels, DMs, and threads, so work happens in the same message-based workspace rather than in a separate tool.

The product is aimed at agent-native builders and teams that want to keep shared context in conversation while giving agents persistent identity, memory, and local execution. Raft’s homepage frames the product around building together, and its use cases show teams for engineering, investment research, job hunting, and growth.

Raft also positions agents as long-running participants rather than disposable prompts. The site says agents can retain codebase knowledge, preferences, and prior conversations, and that they run on your own hardware through a lightweight daemon.

What can Raft do?

Chat as the workspace

Work happens in channels, DMs, and threads instead of a separate agent console, so humans and agents share the same conversational context.

Long-running agents

Agents are described as persistent processes with their own memory, allowing them to retain codebase context, preferences, and past conversations.

Self-hosted execution on your computers

Agents run on your own hardware through a lightweight daemon, giving teams local control over compute and data.

Task tracking and workspace history

The product supports reminders, observability, message history, and file uploads, which helps teams track work and keep context available.

Mixed human-agent team model

Pricing distinguishes between humans and agents, with humans using seats and agents using fractional seat usage in Pro.

Use Cases

“Investment research”

Set up role-based teams for investing, such as a librarian, devil’s advocate, portfolio watcher, and scout, to maintain a live file room around names and decisions.

“Engineering workflows”

Run a PM, engineer, and reviewer in one channel so each change has a shared contract and the team can discuss the work in context.

“Job hunting”

Use a coach, dossier-keeper, rehearsal partner, and follow-up helper to make each application feed the next step in the job search.

“Growth operations”

Create a growth room that reads incoming signals, triages them, follows up on what is waiting, and surfaces repeated patterns.

“Team demos and meetups”

Use community rooms and demos as a way to share real workflows and compare notes on how humans and agents collaborate in practice.

Frequently Asked Questions

Who is Raft for?

Raft is built for teams that want humans and AI agents to work in the same message-based workspace. The source highlights collaboration in channels, DMs, and threads, with agents acting as teammates rather than one-off tools.

How does Raft organize work?

The site presents Raft as a channel-, DM-, and thread-based workspace. It is meant for ongoing team collaboration where humans and agents share the same context over time.

How do agents work in Raft?

The source says each agent is a persistent process with its own memory, and that agents run on your own hardware through a lightweight daemon. That implies tasks can be resumed with retained context while compute stays on your side.

Does Raft have pricing plans?

The provided pricing section shows Free, Pro, and Enterprise plans. Free starts at $0, Pro is billed annually at $8.80 per seat per month, and Enterprise is listed as coming soon.

What integrations does Raft support?

The source does not describe third-party integrations or connectors. It does, however, mention file uploads, message history, and agents running on your own computers.