Getting Started with Barrd
Barrd gives Claude persistent memory across sessions. Connect it once, and every conversation starts with full context — tasks, plans, decisions, and where you left off.
This guide gets you from zero to your first session in under 2 minutes.
Prerequisites
- Claude Code or Claude.ai with MCP support
- A Barrd account — sign up free
Connect Barrd to Claude
Barrd uses the Model Context Protocol (MCP) to communicate with Claude. No plugins or wrappers — just a direct connection.
Option 1: One-click connect
- Log in to your Barrd dashboard
- Click Connect to Claude
- Authorize the connection when prompted
That's it. Claude now has access to your Barrd projects.
Option 2: Manual MCP configuration
If you prefer to configure MCP manually, add Barrd to your Claude MCP settings:
{
"barrd": {
"type": "streamable-http",
"url": "https://barrd.com/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
Get your API key from the dashboard under Settings → API Keys.
Your first session
Once connected, create your first project. You can do this from the dashboard or let Claude do it:
You: Create a new project called "my-app" in Barrd
Claude: Done. Created project “my-app” in Barrd.
I’ll track tasks, decisions, and session
state here as we work.
From this point on, Claude automatically saves context as you work together.
Session memory
This is the core of Barrd. At the end of each session, Claude saves a snapshot of:
- Working memory — what you're currently focused on, what phase you're in
- Current task & plan — so Claude knows what to resume
- Context snapshot — files modified, decisions made, next steps
When you start a new session, Claude loads this automatically:
↳ Loading session state from Barrd...
✓ Resuming from Mar 8: Auth implementation
✓ Current task: MY-APP-12 — Add password reset flow
✓ 3 decisions loaded, 2 tasks remaining
Ready to continue where you left off.
No re-explaining. No lost context. Claude picks up exactly where you left off.
Working with tasks
Barrd includes built-in task tracking that Claude can read and write. Tasks are created automatically as you work, or you can create them explicitly.
You: Let's add authentication to the app
Claude: I’ll create a task to track this.
Created MY-APP-5: “Add authentication”
Let me start with the registration endpoint…
Tasks have statuses (todo, in_progress, blocked, done), can be linked to plans, and carry progress logs with file paths so you always know what changed and where.
Task IDs
Every task gets an auto-generated ID in PROJECT-NUMBER format (e.g., MY-APP-5). Use these IDs to reference tasks in conversation — Claude will look them up instantly.
Plans & decisions
Plans
When work involves 3 or more related tasks, Claude creates a plan — a structured group with a goal, approach, and success criteria. Plans keep complex work organized and give Claude a roadmap to follow across sessions.
Plan: Add authentication system
Status: in_progress (2/4 tasks done)
✓ MY-APP-5 Register + personal org
✓ MY-APP-6 Login + JWT tokens
○ MY-APP-7 Password reset flow
○ MY-APP-8 Two-factor auth
Decisions
Every architectural decision gets logged — what was chosen, why, and what alternatives were considered. Claude references these automatically so you never relitigate the same choices.
Bootstrap your project with CLAUDE.md
The Barrd MCP server includes a built-in CLAUDE.md template that teaches Claude (or OpenCode) how to use Barrd in your project. It covers session management, task tracking, the 3-task planning rule, code intelligence, and more.
Generate it with one tool call:
You: Get the CLAUDE.md template from Barrd and
set it up for this project
Claude: ↳ get_template({name: “claude_md”})
Got it. I’ll customize the template for your
project and save it as CLAUDE.md.
The template includes:
- Session start/end workflows — load and save context automatically
- The 3-task rule — when to create a plan vs. a single task
- Code intelligence setup — AST parsing and LSP integration
- Plan quality guidelines — what makes a good implementation plan
- Test-driven development — write tests first, not after
- Project-specific customization — placeholders for your stack, patterns, and domain terms
There's also a setup guide (get_template({name: "template_guide"})) that walks through customization step by step — which sections to keep, which to remove, and how to adapt examples to your tech stack.
CLAUDE.md file for project instructions.
Teams
On the Team plan, invite your team to share projects, tasks, and context. Everyone — human and AI — works from the same picture.
- Shared projects — all team members see the same tasks and plans
- Shared context — Claude loads team decisions, not just yours
- Org management — invite members, manage roles, SSO
Team setup is available from the dashboard under your organization settings.