Agents & MCP

How it works

Connect an agent in two clicks.

Agents are first-class users, not integrations. Generate a scoped key, drop it into your agent's MCP config, and it's in — operating under exactly the permissions you grant, every action audited.

Generate a scoped key

In your workspace, mint an access token and choose what it can reach — which projects, which actions. Granular by default, least-privilege by choice.

 # Workspace → Settings → Access tokens create_personal_access_token scope:  project:PROD
allow:  create_ticket, transition_state
deny:   archive_*, delete_*
# → kera_pat_9f2a…  (shown once) 

Add it to your agent

Paste the Kera MCP endpoint and key into Claude, Cursor, or any MCP client. That's the whole setup — the agent instantly discovers 150+ tools.

 // .mcp.json { "kera": { "url": "https://kera.eu/mcp/",
"headers": { "Authorization": "Bearer kera_pat_…" } } } # agent discovers 150+ tools 

It works under your rules

The agent acts with exactly the permissions you granted — no more — and every action is audited identically to a human's.

 # The agent can only: ~ act within the scopes you granted
~ on the projects you allowed
# Every action is: ~ audited (who, when, what)
~ versioned · exactly-once

What an agent can do

Anything a person can — bounded by your rules.

The MCP surface mirrors the app. Grant scopes, and the agent operates within them; every action is audited identically to a human's.

Plan & scaffold

Turn a sentence into projects, tickets, dependencies, and owners.

Triage & maintain

Label, assign, transition, and keep the board clean in bulk.

Report from ground truth

Answer "are we on track?" from the audit log and cycle metrics — not a guess.

Drive workflows

React to transitions and events, and take the next action automatically.