A personal work journal

Session

Remember what
you actually worked on.

Session captures your work as structured sessions — so you can recall, reflect, and ask your AI assistant anything about what you've been doing.

Have thoughts? Email us →

The problem

Freelancers forget. And it costs them.

Not because they're disorganised. Because nothing captures work naturally. Notes are scattered. Memory fades. When a client asks what you did last month — or when it's time to invoice — you're rebuilding a story from chat history and calendar entries.

01 "What did I actually work on for this client last month?"
02 "Which of these sessions have I already invoiced?"
03 "Can you summarise what I accomplished this week?"

What a session looks like

Structured by default. Queryable forever.

Every session has a start time, end time, title, tags, and notes. One optional field: project.

Session calendar view with command palette open

09:00 – 10:301h 30m

Gridline authentication

gridlinebackendbugfix

Fixed session timeout issue. Root cause was missing token refresh on idle. Deployed to staging.

13:00 – 15:152h 15m

Northcast episode prep

northcastcontentinvoiced

Prepared AI discussion topics. Outlined episode structure. Shared draft with co-hosts.

How sessions get in

A background analyst, not a form.

Session watches your signals — calendar events, GitHub commits, email threads — and infers what you worked on. At the end of each day it surfaces a short confirmation list. You approve, edit, or dismiss. Nothing is logged without you.

calendargithub · 4 commits

Gridline authentication

09:00 – 11:30 · 2h 30m

calendaremail thread

Client call — Hartwell

14:00 – 14:45 · 45m

At the start and end of each week, Session surfaces a digest — hours logged, projects worked on, uninvoiced time — so you always know where you stand without opening a spreadsheet.

AI interface

Ask anything. Get a real answer.

Because sessions are structured data, any MCP-compatible AI assistant — Claude, ChatGPT, or others — can query your work history naturally. No prompt engineering. No data wrangling.

Claude — connected to Session via MCP

youWhat did I work on for Gridline last week?

claudeYou logged 4 sessions for Gridline last week — 6h 45m total. Main focus was authentication: fixed the session timeout bug and deployed to staging. You also had a 45-minute call on Thursday logged as "client sync."

youWhich of those have I not invoiced yet?

claude3 out of 4 sessions don't have an "invoiced" tag — 6 hours in total. The Thursday client sync is tagged as invoiced.

Principles

Built with constraints.

Personal first

Designed for individuals. No organisations, managers, approvals, or employee monitoring. Ever.

Simple forever

If a feature mainly exists because companies need it, it doesn't belong in Session.

Your data stays yours

CSV and JSON export from day one. No lock-in. No advertising. No selling data.

AI as infrastructure

AI works in two directions: in the background it infers sessions from your signals and asks for confirmation; via MCP, any AI assistant can query what you've built up. Invisible on the way in. Powerful on the way out.

Explicit non-goals

What Session will never be.

Scope discipline is a feature. If you need any of these, use a dedicated tool for it.

× Invoicing

× Payroll

× Teams & org management

× Employee monitoring

× Project management

× Scheduling

× Enterprise reporting

× Accounting

One job. Done well.

A freelancer opens Session, asks their AI assistant what they worked on last month,
and gets a useful answer in five seconds.

This is still an idea we're shaping. Tell us what you think.

Share your feedback