Team chat your AI agents can read.

A team chat for each workspace, a thread for each task and a shared knowledge base, plus an agent that reads all of it. Connect Claude, ChatGPT or any MCP agent and it starts from what the team already decided, not from a blank prompt.

Free for teams up to 3 · Claude connects with no API key

Acme
Onboarding wizard
Safari login loop
Pricing page copy
Acme Labs
Ask Scindo
Help
Settings
Onboarding wizard· Acme4
M

mara · 10:12

Do invites need to happen before the first channel is created?

J

jin · 10:14

No. Move them to the end of the wizard, people skip them anyway.

M

mara · 10:15

Works for me. Let's cut the team-size step too, nobody fills it in.

Scindo · 10:15 · agent

Saved two decisions from this thread and updated the plan. Anyone who asks an agent about onboarding will get these from now on.

Decision

Invites move to the end of the wizard

Onboarding wizard · 10/24/2026 · Jin

Decision

Cut the team-size step

Onboarding wizard · 10/24/2026 · Mara

Plan updated · 2 of 5 steps
Message your team or @agent...Scindo

Works where you already build

Claude
Gemini CLI
ChatGPT
Grok
Copilot
VS Code
CLI
GitHub
Jira
Linear
Notion
Google Docs
Figma

The problem

AI agents can write code. They don’t know what your team decided.

Every agent session starts from zero. The decisions live in a chat thread, the plan lives in someone’s head, and the PR surprises the team. Scindo keeps the agreement in one place and hands it to every agent that asks.

01

Plans live in someone's head

PM explains in Slack. Developer builds from memory. Details don't make it.

02

AI is single-player

Everyone has their own AI session. Nobody sees what others discussed. Same questions, asked again.

03

PRs surprise the team

“That's not what we agreed on.” Rework. Sprint slips. Every week.

What agents can see

Everything your team agreed. Now your agents can read it too.

Decisions, plans, threads, tasks, documents and pull requests. Searchable by any MCP client, scoped to exactly what you can see. Direct messages never included.

Decision

Cut the team-size step from the wizard

Onboarding wizard · 10/24/2026 · Jin

Decisions

Saved from chat in one click

Agents read the decision, who made it, and the thread it came from.

Plan · Billing page v2

  1. Invoice list
  2. Usage meter
  3. 3Plan switcher
  4. 4Empty states

Plans

The agreement, step by step

Drafted from the thread, approved by the team, read by the developer and the agent.

M

Do invites need to happen before the first channel?

J

No. Move them to the end, people skip them anyway.

A

Noted. I'll draft the plan with invites last.

Threads

The conversation that led there

Edge cases, pushback and the final call, in the order they happened.

Billing page v2
In progress
Members pane
Blocked
Search across workspaces
In review

Tasks

What is in flight, and what is blocked

Status, owner, latest message. Agents answer “what are we waiting on?” honestly.

Data access decisions

Documents

The knowledge base, in plain Markdown

Specs, runbooks, imported Notion pages. Searchable and readable in full.

#214 Billing page v2
CI passed

Pull requests

Plan next to the diff

The reviewer, and the agent, see what was agreed beside what was built.

Ask anything

Ask in plain words. Get cited answers back.

Click a prompt to run it.

Scindo MCP · works with any MCP client
Y

You

What did the team decide about onboarding last week?

Answers cite the thread, task or document they came from

When to reach for it

Three moments where a blank prompt costs you a day.

Start from the plan, not from memory

Your agent reads the approved plan and the thread behind it before writing a line. Goals, edge cases and the decisions already made are in its context from the first prompt.

Agent · Claude Code

  • Read plan: Onboarding wizard v2
  • Pull decisions from the task thread
  • Scaffold /onboarding route
  • Drop team-size step
  • Open PR with plan linked

Setup

Set up in under a minute.

One command for Claude Code, Gemini CLI or Codex, one connector for Claude. The endpoint is the same for all of them.

Step 1

Create an API key

Settings → API Keys. The key sees only what you can see: your channels, workspace documents and your own private ones. Revoke it any time.

Key name

Claude Code · laptop

All workspaces
Free plan

Step 2

Connect your tool

Pick the agent you use. Claude needs no key at all.

No API key. Settings → Connectors → Add custom connector, paste this URL and sign in.

https://app.scindo.one/api/mcp

Pricing

Per workspace, not per seat.

Free

Free forever

$0

  • 1 workspace
  • Up to 3 members
  • 50 AI responses/mo
  • Knowledge base + Ask
  • Scindo MCP
  • 1 GitHub repo
  • VS Code + CLI

Team

Per workspace / month

$29

  • Unlimited members
  • 1,000 AI responses/mo
  • All 5 agent roles
  • 3 GitHub repos
  • 1 GB storage
  • GitHub integration

Pro

Per workspace / month

$79

  • Unlimited AI responses
  • Unlimited repos
  • Smart model routing
  • BYOA + external MCP servers
  • 5 GB storage
  • Webhooks + analytics

FAQ

Questions.

What is Scindo MCP?

Scindo exposes your workspace as an MCP server. Any MCP client, such as Claude, Claude Code, Gemini CLI or Codex, can read channels, search messages, list tasks, query the knowledge base and save decisions back to it.

What can a connected agent see?

Exactly what you can see: public channels, private channels you belong to, task threads, workspace documents and your own private documents. Direct messages are never included. Revoke the key and the agent is disconnected.

Which tools work with it?

Claude (web, desktop, mobile) via a connector with no key. Claude Code, Gemini CLI, Codex and any client that accepts an HTTP MCP server with headers via an API key. Claude Desktop through mcp-remote.

Is it free?

Yes. Scindo MCP is on every plan, including Free. Connecting your own external MCP server to Scindo's agent is a Pro feature.

Why per workspace, not per seat?

Everyone should be in the conversation, not just engineers. Per-seat pricing makes teams leave people out. We don't want that.

What counts as an AI response?

One reply from Scindo's agent in a channel, task, or Ask Scindo. Tool calls inside that reply don't count separately. Questions your own AI tool asks through Scindo MCP don't count: that reply comes from your tool, not from Scindo's agent.