MCP server for your agents

A new AI data layer to capture time & memory.

Less context. Fewer tokens. Better continuity — for you, your agents, and everyone you work with. No screenshots, no software installed.

Create free account

The demo opens a real, read-only workspace — no signup, no email. Takes a few seconds.

Claude CodeCodexClaude.ai connectorMCP + REST APIProjectsRoles & invites

Pulse Engineer mode: a week of entries with an activity heatmap, a breakdown by category, and rows logged by an AI agent.

Engineer mode: your activity and this week's entries. Rows marked [agent] were logged automatically by your AI while it worked.

Agents × people

It starts solo. It becomes memory.

Solo

One person's work, agent captured your work instead of you — no timer, no manual entry.

Multi-agents

Claude Code, Codex, and every other agent write to the same record instead of starting cold.

Team

Invite teammates and the record grows with them — everyone keeps their own, visible to admins.

Memory

What's left behind isn't a timesheet. It's shared context every agent can read from.

Pricing

Two ways to start

Solo

ReadyFree while in beta

Connect Claude Code or Codex via MCP — every session becomes a structured record. Timer and manual entry work too.

Small

ReadyFree while in beta

Invite teammates by email and set their role and mode. Admins can open any member's record read-only.

Both demos open a shared, read-only workspace with realistic data — no signup, no email. Pricing is still being finalized; no credit card required until then, and we'll give notice before anything changes.

How it works

Three steps, then it runs itself

Connect your agent

Create a token in Settings → Developers and point Claude Code or Codex at the MCP server — or connect from Claude.ai as a custom connector over OAuth.

Just work

Your agent logs entries as it goes — duration, note, category, and project — through the MCP tool. Nothing to start, nothing to stop, nothing to remember.

Review and own it

Each agent entry is badged so you can edit, re-attribute, or delete it. Export your whole work record as JSON whenever you want.

What's inside

Everything a work record needs

Agent-logged

Records that write themselves

Claude Code and Codex log entries as you work — badged [agent], no timer required.

API + MCP

MCP-native, REST when you need it

A REST API plus a 13-tool MCP server — 19 with work memory switched on — for entries, summaries, and more.

Your workspace

Roles, modes, and a settings screen

Invite by email; set roles and modes per person, and manage projects and categories.

The pulse-capture skill in Claude Code: proposed time entries reconstructed from session history, with durations, categories and projects, a 796-minute total, and the entries logged to Pulse once approved.

Developers

An MCP server and a REST API, not a plugin marketplace

The agent is the integration layer. Pulse builds no GitHub, Linear, Slack, or Calendar connectors — it turns work into records through the connections your agent already has. Point any MCP client at the remote server:

{ "pulse": {
    "url": "https://pulse.transmit.work/api/mcp",
    "headers": { "Authorization": "Bearer psk_..." }
} }

Log and edit from the agent

pulse_log_entry and pulse_update_entry write to the caller's own work record — with an optional project label in the same call.

Read what you need

Whoami, entries, summaries, categories, tools, and projects, plus member lists for admins — the same data the UI reads.

Remember what mattered

Switch it on per workspace: when your agent notices a decision or a lesson, it proposes a work memory. You confirm the ones worth keeping, and future agents read them back.

REST when MCP isn't the fit

Bearer-token endpoints under /api/v1 cover the same ground for scripts, cron jobs, and dashboards.

Full API reference — every MCP tool and REST endpoint →

Privacy by design

What Pulse stores, and what it never sees

No screen, no keystrokes

No screenshots, keystrokes, window titles, idle tracking, browsing, or location — ever.

No software installed

Pulse is an API your agent calls — nothing installed on your machine, nothing running in the background.

Agents can't delete

No delete tool, by design — every entry your agent writes stays visible and editable.

AI tools have context. Give it a record.

Free to try while we're in beta.

Create free account

Enterprise

Your memory. Your infrastructure.

For organizations that need Pulse inside their own security boundary. Pulse can be deployed around your requirements — including your own infrastructure, database, and authentication — while keeping the same AI-native work and decision-memory experience.

Infrastructure

Pulse-managed, dedicated, or your environment.

Data

Dedicated or customer-controlled database.

Auth

Integrate with your identity and access policies.

Deployment

Cloud, dedicated, or internal options depending on requirements.

Ask about Enterprise →