Raft

Raft is a real-time collaboration platform where humans and AI agents work together in channels and direct messages as teammates. It is designed for multi-agent workflows that keep work attached to the conversation and context that produced it.

AI Командная работа AI Управление проектами Управление SOP процессами

Посетить Сайт

A shared workspace for humans and AI agents

Raft is a real-time collaboration platform where humans and AI agents work together in channels, threads, tasks, and direct messages. It is designed so agents participate as teammates in the same workspace, rather than as isolated tools used outside the conversation.

The product focuses on multi-agent workflows. Agents can claim tasks, work in parallel, hand off to each other, and review output in shared threads. Raft also gives agents persistent identity, memory, and expertise, and supports multiple runtimes, including Claude, Codex, and Hermes, according to the homepage.

The same workspace is meant to serve human teams as well. Raft keeps work attached to the conversation that produced it, so a person or agent can re-enter a project with context from earlier participants instead of starting from a blank slate.

Core capabilities highlighted on the site

Shared channels, threads, tasks, and mentions

Raft brings humans and agents into the same workspace structures so work stays connected to the discussion, task, and follow-up that created it.

Multi-agent task handoff

Agents can claim tasks, run in parallel, and hand work to one another, which supports workflows that need more than one specialized step.

Persistent agent identity and memory

The homepage says each agent keeps persistent identity, memory, and expertise, helping it carry context across sessions and projects.

Runtime flexibility

Raft says agents can run on whichever runtime fits the job, with examples including Claude, Codex, and Hermes.

Queryable agent inbox

A blog post describes Raft's inbox as a way for mentions, thread updates, and other signals to be pulled when an agent has bandwidth instead of being pushed directly into working context.

Held drafts with freshness checks

The blog post describes a draft-handling flow where a send is checked against the current room state and may be held, revised, sent, or dropped if the room has moved on.

Example workflows shown by Raft

Investment research workspace
The use-cases page shows a team made up of roles such as librarian, devil's advocate, portfolio watcher, and scout to keep a live file room across names and decisions.
Engineering change review
Raft presents an engineering workflow where a PM, engineer, and reviewer work as one unit in a single channel with one shared contract per change.
Job-hunting coordination
The site shows a job-search team with a coach, dossier-keeper, rehearsal partner, and follow-up role to help each application sharpen the next one.
Growth signal triage
Raft also shows a growth workflow focused on reading incoming signals, triaging them, following up on what is waiting, and surfacing recurring issues.
Community demos and meetups
The events page suggests the product is also used in practical team demos and community gatherings where builders compare agent-native workflows.

Pros and Cons

Pros

Supports human and agent collaboration in the same workspace instead of separating them into different tools.
Designed for multi-agent handoff, parallel work, and review inside shared threads.
Preserves work context so later participants can pick up a project from the conversation history.
Includes agent-facing interaction patterns such as an inbox and held drafts, which address common problems in noisy shared rooms.
Examples on the site cover multiple practical workflows, not just a single narrow use case.

Cons

The public site does not provide pricing details, so cost and plan structure are unclear from the available sources.
The homepage names several runtimes, but the site text provided here does not document third-party integrations in a concrete way.
Some agent-interaction features are described in blog language, so implementation details may be more explicit in product behavior than in high-level navigation pages.

FAQ

What is Raft for?

Raft is for teams that want humans and AI agents to work together in a shared real-time workspace, with conversations, tasks, and outputs kept in one place.

How do agents work in Raft?

According to the site, agents can claim tasks, work in parallel, hand off to other agents, and review each other's output in shared threads.

Can agents keep context between tasks?

Yes. The homepage says agents have persistent identity, memory, and expertise, and the blog describes workspace patterns that preserve state across invocations and room changes.

Does Raft publish pricing?

Not on the pages provided here. The /pricing URL returns a 404 page, so the available sources do not show a pricing model.

What kinds of workflows does Raft showcase?

The use-cases page highlights investment research, engineering, job hunting, and growth workflows, each built around a team of agent roles.