How-to

Create an agent and set up Claude Code

Create a workspace agent, then prepare the Claude Code connection it will use.

2 min 9 sFor: Pilot developer10 steps

Read the steps OR watch the video. Both teach the same outcome, and neither is needed for the other.Jump to the video

Quick steps

  1. Click SettingsStart on DM Home, in your workspace. In the sidebar, open Settings.
    1. Start on DM Home (/).
    2. Click "Settings". The next view is "Workspace settings" (/settings/workspace).
    3. You should now see "Agents".

    What it means: Start on DM Home, in your workspace. In the sidebar, open Settings.

  2. Click AgentsSettings holds the shared setup for your workspace. Open Agents.
    1. Click "Agents". The next view is "Agents" (/control-panel/agents).

    What it means: Settings holds the shared setup for your workspace. Open Agents.

  3. Click Create agentThis is where DM tracks agents such as Claude Code. None exist yet, so select Create agent.
    1. Click "Create agent". The next view is "Create workspace agent" (/control-panel/agents).
    2. You should now see "Display name".

    What it means: This is where DM tracks agents such as Claude Code. None exist yet, so select Create agent.

  4. Fill in the formEnter Claude Code as the display name, with an optional description. DM turns the name into a stable agent key. It identifies the agent and is not a secret.
    1. Type "Claude Code" into the "Display name" field.
    2. Type "Coding agent that searches approved decisions before changing code." into the "Description (optional)" field.
    3. You should now see "Claude Code".

    What it means: Enter Claude Code as the display name, with an optional description. DM turns the name into a stable agent key. It identifies the agent and is not a secret.

  5. Review default access profileReview the access profile. Standard is the default. A Standard agent reads what it may see and proposes changes for review. It cannot approve anything. Restricted gives less access, and Pilot Full Access gives more.
    1. Review "Default access profile": "Standard Read access to authorized shared knowledge; may submit proposed changes for revi…".

    What it means: Review the access profile. Standard is the default. A Standard agent reads what it may see and proposes changes for review. It cannot approve anything. Restricted gives less access, and Pilot Full Access gives more.

  6. Click CreateSelect Create. DM saves the agent and opens its details page. The Connected badge only means DM is prepared. Claude Code is not set up yet.
    1. Click "Create". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001).
    2. You should now see "Agent details": "Connected Workspace Agent Claude Code Coding agent that searches approved decisions befor…".

    What it means: Select Create. DM saves the agent and opens its details page. The Connected badge only means DM is prepared. Claude Code is not set up yet.

  7. Review connections for this agentMCP Connections lists what this agent can reach. Anything your plan excludes shows as unavailable.
    1. Review "Connections for this agent": "MCP connections for this Agent This setup uses one AuzzurA Agent installation and include…".

    What it means: MCP Connections lists what this agent can reach. Anything your plan excludes shows as unavailable.

  8. Click Copy OAuth configurationThe recommended method is Connect with authentication. It needs no API key. Copy the OAuth configuration. It holds your DM MCP URL and no secret. Add that URL to Claude Code. On first connection, your browser opens so you can sign in and review access.
    1. Click "Copy OAuth configuration". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001).
    2. You should now see "Copied".

    What it means: The recommended method is Connect with authentication. It needs no API key. Copy the OAuth configuration. It holds your DM MCP URL and no secret. Add that URL to Claude Code. On first connection, your browser opens so you can sign in and review access.

  9. Click Claude CodeIf you cannot sign in through a browser, use the Claude Code tab for manual credentials. Its configuration holds placeholders only. DM shows a new secret once, so keep it private.
    1. Click "Claude Code". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001).
    2. You should now see "Placeholders only": "{ "mcpServers": { "decision-memory": { "url": "https://dm.example.test/mcp", "headers": {…".

    What it means: If you cannot sign in through a browser, use the Claude Code tab for manual credentials. Its configuration holds placeholders only. DM shows a new secret once, so keep it private.

  10. Click Test connectionTest connection checks the setup from DM's side. In this demo the result is simulated. In your workspace, ready means DM is prepared, not that Claude Code has connected.
    1. Click "Test connection". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001).
    2. You should now see "Simulated results": "initialize ready tools/list ready installation active grants ready".

    What it means: Test connection checks the setup from DM's side. In this demo the result is simulated. In your workspace, ready means DM is prepared, not that Claude Code has connected.

Next, add DM in Claude Code, complete the sign-in, then check this agent's runs in DM.

Statuses such as the Connected badge and the Test connection results were simulated in the recording. Adding DM in Claude Code, the browser sign-in and a real tool call were not verified.

Descriptive transcript

Read the full transcript, including what happens on screen

Demo environment · synthetic workspace · statuses are simulated. Demo walkthrough of the DM-side setup · Settings → Agents → MCP Connections
This guide shows the DM-side setup only. It does not show Claude Code being connected or authenticated.

A banner reading "Demo environment · synthetic workspace · statuses are simulated" stays on screen for the whole video.

00:00–00:07 · intro

On screen: The AuzzurA DM logo, the title "How to Set Up a Claude Code Connection to AuzzurA DM" and the subtitle "Demo walkthrough of the DM-side setup · Settings → Agents → MCP Connections".

Narrator: This demo shows how to set up a Claude Code connection to AuzzurA DM, using a workspace with simulated data.

00:07–00:15 · nav-home

On screen: Start on DM Home (/). Click "Settings". The next view is "Workspace settings" (/settings/workspace). A highlight marks "Settings". Afterwards: A highlight marks "Agents".

Narrator: Start on DM Home, in your workspace.

Narrator: In the sidebar, open Settings.

00:15–00:22 · nav-settings

On screen: Click "Agents". The next view is "Agents" (/control-panel/agents). A highlight marks "Agents".

Narrator: Settings holds the shared setup for your workspace.

Narrator: Open Agents.

00:22–00:30 · open-create

On screen: Click "Create agent". The next view is "Create workspace agent" (/control-panel/agents). A highlight marks "Create agent". Afterwards: A highlight marks "Display name".

Narrator: This is where DM tracks agents such as Claude Code.

Narrator: None exist yet, so select Create agent.

00:30–00:41 · create-name

On screen: Type "Claude Code" into the "Display name" field. Type "Coding agent that searches approved decisions before changing code." into the "Description (optional)" field. A highlight marks "Name the agent". Afterwards: A highlight marks "Claude Code".

Narrator: Enter Claude Code as the display name, with an optional description.

Narrator: DM turns the name into a stable agent key. It identifies the agent and is not a secret.

00:41–00:57 · create-profile

On screen: A highlight labelled "Default access profile" marks "Standard Read access to authorized shared knowledge; may submit propo…".

Narrator: Review the access profile. Standard is the default.

Narrator: A Standard agent reads what it may see and proposes changes for review. It cannot approve anything.

Narrator: Restricted gives less access, and Pilot Full Access gives more.

00:57–01:09 · create-submit

On screen: Click "Create". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001). A highlight marks "Create". Afterwards: A highlight labelled "Agent details" marks "Connected Workspace Agent Claude Code Coding agent that searches appr…".

Narrator: Select Create. DM saves the agent and opens its details page.

Narrator: The Connected badge only means DM is prepared. Claude Code is not set up yet.

01:09–01:16 · connections

On screen: A highlight labelled "Connections for this agent" marks "MCP connections for this Agent This setup uses one AuzzurA Agent inst…".

Narrator: MCP Connections lists what this agent can reach. Anything your plan excludes shows as unavailable.

01:16–01:36 · oauth-copy

On screen: Click "Copy OAuth configuration". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001). A highlight labelled "Recommended - no API key" marks "Connect with authentication Connected Add the DM MCP URL to your Agen…". Afterwards: A highlight marks "Copied".

Narrator: The recommended method is Connect with authentication. It needs no API key.

Narrator: Copy the OAuth configuration. It holds your DM MCP URL and no secret.

Narrator: Add that URL to Claude Code. On first connection, your browser opens so you can sign in and review access.

01:36–01:49 · manual-fallback

On screen: Click "Claude Code". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001). A highlight labelled "Fallback - manual credentials" marks "Claude Code". Afterwards: A highlight labelled "Placeholders only" marks "{ "mcpServers": { "decision-memory": { "url": "https://dm.example.tes…".

Narrator: If you cannot sign in through a browser, use the Claude Code tab for manual credentials.

Narrator: Its configuration holds placeholders only. DM shows a new secret once, so keep it private.

01:49–02:02 · test-connection

On screen: Click "Test connection". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001). A highlight marks "Test connection". Afterwards: A highlight labelled "Simulated results" marks "initialize ready tools/list ready installation active grants ready".

Narrator: Test connection checks the setup from DM's side. In this demo the result is simulated.

Narrator: In your workspace, ready means DM is prepared, not that Claude Code has connected.

02:02–02:09 · outro

On screen: The AuzzurA DM logo, the heading "Next: connect from Claude Code" and the line "Add the DM MCP URL, sign in, and review access in your browser."

Narrator: Next, add DM in Claude Code, complete the sign-in, then check this agent's runs in DM.

Last verified against the product on .

How-to 2 min 8 s10 steps

Prepare relevant context for a task

Create a bounded context package from approved knowledge. Before your AI agent changes code, give it the decisions that apply.

For: Coding-agent user

How-to 1 min 54 s10 steps

Find why a decision was made

Retrieve rationale, evidence and change history without chasing old messages. You found the implementation. But why was that choice made?

For: New engineer or tech lead

How-to 1 min 20 s9 steps

Review and approve a decision candidate

Demonstrate why AI extraction is not automatically team policy. DM proposes what it found. Your team decides what becomes trusted.

For: Reviewer / workspace decision owner

How-to 49 s7 steps

Add your first evidence to DM

Start from one relevant source rather than recording everything. A team decision often starts in a document, a ticket, or a meeting note.

For: Workspace member