Connect the Pulse MCP Server

Connect an AI agent — Claude, Cursor, or any Model Context Protocol client — to your Pulse workspace, and it can plan sprints, triage the backlog, log risks, and report on delivery health for you. Everything the agent does is checked against your own permissions: it can only see and change what you can.

Quick start

Connecting takes two steps: create a personal API key, then add the server to your client.

1. Create an API key

In Pulse, open Profile → API Keys and click Create API key. Give it a name, pick the scopes it should have, and choose an expiry (30, 90, or 365 days — or never). Copy the key when it appears: it starts with pulse_sk_ and is shown only once.

2. Add the server to your client

MCP is an open standard, so every client connects the same way. Whatever tool you use, it needs the same three things:

• Endpoint — https://api.pulseios.com/mcp
• Transport — HTTP (streamable)
• Header — Authorization: Bearer pulse_sk_YOUR_KEY

Pick your client below and paste your key in place of pulse_sk_YOUR_KEY.

Claude Code

One command in your terminal:

Cursor

Add this to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (this project only), then reload Cursor:

VS Code (GitHub Copilot)

Create .vscode/mcp.json in your project — or run “MCP: Add Server” from the Command Palette. Note VS Code uses the key servers, not mcpServers:

Windsurf

Add this to ~/.codeium/windsurf/mcp_config.json, then hit Refresh in the Cascade MCP panel. Windsurf uses serverUrl for remote servers:

Any other MCP client

Most clients — Cline, Zed, JetBrains AI Assistant, and others — take a standard JSON config. Point it at the HTTP endpoint and pass the header:

stdio-only clients (Claude Desktop, Codex, Gemini CLI)

Some clients only speak stdio and can’t call a remote URL directly. Bridge to Pulse with mcp-remote — npx fetches it on first run, nothing to install:

OpenAI Codex uses the same command in TOML (~/.codex/config.toml):

That’s it — reload your client and your agent is connected to your Pulse workspace.

Permissions & security

• A key acts as you: its scopes are capped to your workspace role, and every tool call is also checked against your live role-based permissions (RBAC).
• Grant the minimum: scopes are split into read, write, and delete per domain, so a reporting agent can be read-only.
• Revoke any time from Profile → API Keys — revocation and expiry take effect on the very next request.
• Treat pulse_sk_ keys like passwords: store them in a secret manager and never commit them to a repository.
• A key is bound to one workspace. To work across workspaces, create a key in each.

What your agent can do

The server exposes roughly 110 tools across the following domains. Read scopes are separate from write and delete, so you can grant an agent exactly the access it needs.

Tasks

• list-tasks — list and filter tasks by status, assignee, priority, or search (plus overdue / blocked presets)
• get-task — full task details with recent inline comments
• create-task, update-task, delete-task — manage individual tasks
• assign-task, change-task-status — reassign and move tasks through the workflow
• bulk-create-tasks, bulk-update-tasks — batch changes (up to 20 at a time)
• add-task-comment — comment on a task with markdown and source attribution

Tags

• list-tags, list-task-tags, list-tasks-by-tag
• add-tag-to-task, remove-tag-from-task (auto-creates tags as needed)

Projects

• list-projects, get-project (optional cancellation what-if simulation), update-project
• get-project-health — combined red/yellow/green from RAID, tasks, milestones, readiness
• get-readiness-history, get-project-status-history — chronological score and phase changes

Sprints

• list-sprints, get-sprint, create-sprint, update-sprint

Milestones

• list-milestones, get-milestone, create-milestone, update-milestone
• bulk-update-milestones (up to 20)
• list-decision-log — chronological go/no-go decisions across milestones

Members & workload

• get-current-user, list-members, get-member
• get-member-workload — tasks, story points, allocation %, overdue, blocked, skills per person

Goals & OKRs

• list-goals, get-goal, get-goal-health
• list-okrs, get-okr — OKRs with key results and optional health / strategy alignment
• get-okr-driver-attribution — rank the projects driving an OKR or goal

Workspace & analytics

• get-workspace, get-workspace-health, get-workspace-tasks-status
• get-team-workload, get-velocity-metrics, get-velocity-forecast (p50/p80/p95)
• get-project-analytics, get-milestone-timeline, get-risks-summary
• get-capacity-burn-rate, get-story-points-coverage

Capacity & ownership

• list-resource-allocations, get-resource-conflicts — spot over-allocation across overlapping windows
• get-ownership-concentration — detect single points of failure across goals, OKRs, projects, tasks

RAID

• list-raids, get-raid, create-raid, update-raid, delete-raid

Strategy & dependencies

• get-cross-project-dependency-graph — workspace-wide dependency graph with cycle detection
• list-strategy-artifacts, get-strategy-artifact — vision, mission, strategic themes
• request-pulse-analysis — re-analyze real delivery state after agent work

Documents & files

• list-documents, create-document, update-document, publish-document, delete-document, move-document
• list-adrs, create-adr — Architecture Decision Records
• read-file — extract text from uploaded PDFs, Office files, and images

Scheduled workflows

• list-workflows — see your scheduled AI automations and their status
• upsert-workflow — create or update a scheduled automation (preset, cron, or interval schedule)
• run-workflow-now, set-workflow-status — trigger a run immediately, pause or resume

Integrations

• GitHub — list repos, branches, and PRs; create branches and PRs; link a branch or PR to a task
• Jira — search issues; list projects, issue types, statuses, and sprints
• Linear — search issues; list projects and teams
• Asana — search tasks; list projects and users
• Trello — search cards; list boards and lists
• Monday.com — search items; list boards and users
• HubSpot CRM — deals, contacts, companies, and a rolled-up CRM summary

Resources & prompts

Beyond tools, the server publishes read-only resources under the pulse:// URI scheme — including workspace projects, per-project sprint boards and backlog trees, and a workspace health summary. It also ships reusable prompt templates: standup-summary, triage-backlog, sprint-planning, and risk-assessment.

Claude Code skill

Pulse ships a Claude Code agent skill, pulse-mcp, that teaches the agent how to use these tools well: which tool to reach for, when to batch, how IDs and human keys work, and the gotchas that trip agents up. Drop the skill folder into ~/.claude/skills/ (or your project’s .claude/skills/) and Claude picks it up automatically whenever you work with your Pulse workspace.

• SKILL.md — connection guide, tool map, entity model, best practices, and gotchas
• references/tool-catalog.md — every tool with its parameters, scopes, and required permissions
• references/playbook.md — how to run a project well through the MCP surface, from backlog to delivery health

Ready to go? Create your key under Profile → API Keys and connect your first agent in under a minute.