Now in early access · VS Code · macOS + Windows

Simmer remembers why
the code looks the
way it does.

Every AI tool knows what your code is.

None know why it became that way, or where you left off.

Simmer does. Because it was there when the code was written.

// no spam  ·  early access when we launch

you're on the list. we'll reach out soon.
See how it works ↓
auth/middleware.ts
Re-entry brief Yesterday · 4:52pm · 2h 15m
Debugging the token refresh race condition in the auth middleware.
Ruled out
✓Middleware layer is not the source
✓Token expiry logic confirmed correct
Next step
Check callback timing in refreshToken.ts
How it works

Three things.
One loop.

The VS Code extension runs silently. You do nothing differently. Simmer has a richer picture of your intent than any commit message ever captured.

01
Simmer watches
The VS Code extension runs silently. It observes file edits, stall patterns, git activity, and AI agent sessions: every prompt, every output, every decision you made reviewing it. You do nothing differently.
02
Simmer remembers
Every decision, every dead end, every vibe-coded session. All stored locally. After a few days, Simmer has a richer picture of your intent than any commit message ever captured.
03
Simmer surfaces
A quiet tap on the shoulder from the system tray. Not a chatbot. Not a notification wall. One specific thing, at the right moment, grounded in your own history.
Getting started

Useful from day one.
Indispensable by week four.

Your repo has months of history. Simmer reads it on install. The baseline was already there — you just hadn't read it yet.

D1
Day one
Install. Simmer reads your history.
On install, a one-time background job scans your git history. No prompt, no setup, nothing visible. Within minutes, Simmer has co-edit maps, churn signals, ownership patterns, and 90 days of commit intent — before you've written a single new line.
Git baseline, day one auth/ + middleware/ committed together 78% of the time
payments/charge.ts: highest revert rate in the repo
Alex: 23 of last 30 auth/ commits — likely owner
D3
Day three
Your first re-entry brief, grounded in history
After your first multi-hour session, Simmer has live behavioral data to work with — and it already had the git baseline. Your first re-entry brief isn't just the last 30 minutes. It knows which files you've historically circled back to, and where you've gotten stuck before.
First brief, typical "Yesterday at 4:52pm you were debugging the token refresh race condition in auth/middleware.ts. You ruled out the middleware layer. Next step: check callback timing in refreshToken.ts."
W2
Week two
Stall detection activates
With a week of live session data, Simmer can now cross-reference behavioral signals against the churn patterns it already read from git. When you get stuck on a file with a high historical revert rate, Simmer already knows this is a likely stall zone. The first time it surfaces a fix you wrote weeks ago for a problem you're hitting today, you'll feel it.
What changes Stall nudges appear in the system tray. One suggestion. Specific. Grounded in your code, not generic Stack Overflow advice.
W4
Week four
Handoff summaries and AI context
A month of session data is enough to generate your first full handoff summary, either when a ticket changes hands or when Simmer detects you haven't touched a file in a while. The AI context injection layer means Claude or Copilot now have your recent session history automatically.
The unlock A teammate picks up your ticket and sees not just the code, but what you were attempting, what you tried, and where you got stuck — without you writing a word of documentation.
M3
Month three onward
Irreplaceable history
Three months of behavioral telemetry is a data asset that cannot be reconstructed retroactively. The gap between what your codebase does and why it looks the way it does is fully closed, for every developer on the team, for every file, for every AI-generated session.
Three problems

The thinking disappears.
Simmer keeps it.

Context switching

You were deep in it yesterday. Today you're spending 20 minutes just remembering where you were. That's not a you problem. It's a tooling gap. Simmer closes it.

Team handoffs

A teammate picks up your ticket. The code is there. The thinking behind it isn't. Simmer was capturing it the whole time and surfaces it automatically when they open the file.

Vibe coding memory

Your AI agent wrote 400 lines last Thursday. Three weeks from now, someone needs to extend it. They'll have no idea why it looks the way it does. Simmer does. It was in the room.

Why different

Not another AI tool
that waits to be asked.

Cursor knows your code. Simmer knows your team.

Others Simmer
Knows what your code is ✓ ✓
Knows why it became that way ✗ ✓
Was present during the AI session ✗ ✓
Remembers where you left off ✗ ✓
Generates handoff briefs automatically ✗ ✓
Requires you to do extra work ✓ ✗
Starts fresh every session ✓ ✗

Captured as a byproduct of work, not as extra work.

Built to scale

From you to your
whole team.

Starts as your personal re-entry layer. Scales to your whole team.

  • Why that architectural choice was made six months ago
  • When two engineers are stuck on the same problem
  • Instant context for every new hire, on every file

Make your team's thinking visible.

Solo
Re-entry briefs · Stall intervention · Local only · Nothing leaves your machine
Team
Shared session context · Automatic handoff briefs · PR intent layer · "who knows this file" routing · Linear/Jira integration
Org
New hire onboarding mode · Codebase intent maps · Cross-team pattern insights · SSO + on-prem option
Developers from these communities are already waiting →
Reddit Hacker News Dev.to X (Twitter)
Privacy architecture

Built local-first.
By design, not by promise.

Simmer observes more about how you work than any tool you've used. That means the architecture choices matter. Here's exactly how we made them.

Data flow
Your machine — all behavioral telemetry Keystrokes, file edits, stall patterns, AI prompts, session records, decision traces. Processed on-device. Never transmitted.
↓ local processing only ↓
Local model — intent inference The model that turns raw telemetry into re-entry briefs and stall nudges runs entirely on your machine. No cloud call. No latency.
↓ explicit export only, team plan ↓
Team sync — opt-in, session summaries only On team plan, you choose what to share: summarised session records (not raw telemetry). You control what leaves the machine.
↓ never ↓
Anthropic / any third party Raw telemetry, keystroke data, or anything that identifies individual developer behavior is never transmitted to us or anyone else.

What Simmer captures

→File edit patterns — which files you edit together and how often
→Stall signatures — where you get stuck and for how long
→AI session context — prompts you sent, outputs accepted or edited
→Decision traces — what you tried, reverted, then tried again
→Session timestamps — when you picked up a task and when you left it
Never captured or transmitted
  • The content of your code (Simmer reads structure, not source)
  • Credentials, secrets, or environment variables
  • Individual developer identity to your manager or org (without consent)
  • Raw keystroke or mouse telemetry
  • Anything to any third party, ever

Managers don't see individual data

Team plan surfaces shared session context and handoff summaries. It does not expose individual stall patterns, error rates, or session durations to managers. Developer data is for developers. Org-level insights are structural: patterns, not individuals.

Pricing

Simple pricing.
No surprises.

Individual
Free during beta
  • Stall detection + intervention
  • Re-entry briefs
  • Local decision memory
  • VS Code extension + system tray app
  • Nothing leaves your machine
Join waitlist
Most popular
Team
Coming soon ~$20/dev/mo
  • Everything in Individual
  • Team session sync
  • Automatic handoff briefs
  • PR context layer
  • Linear / Jira / Slack integration
Notify me
Enterprise
Contact us
  • Everything in Team
  • SSO + admin controls
  • On-prem / air-gapped
  • SLA + dedicated support
  • Custom integrations
Get in touch
Design partners

Building this with
real teams.

We're working with a small number of engineering teams before launch. Not a pilot — a genuine co-build. Your team's real pain shapes the product.

What design partners do

  • Use Simmer with 3–10 engineers for 6–8 weeks on real production work
  • Weekly 30-min call with the founding team to tell us what's working and what isn't
  • Share anonymised patterns (with consent) to help us calibrate stall detection and handoff briefs
  • Give honest feedback. If it's not useful, say so.

What you get

  • Early access, free for the duration of the design partnership
  • Direct line to the founders; feature requests go straight to the roadmap
  • Locked-in team pricing when we launch ($12/dev/mo vs ~$20)
  • First-named reference if the product solves your problem

Apply to be a design partner

We're selective. We want teams actively using AI coding tools and a codebase complex enough that context actually matters. 5–50 engineers.

application received — we'll be in touch.

Be the first to never
lose your place.

Simmer is in private development. Join the waitlist for early access when we launch.

// free during beta  ·  no credit card

you're on the list — we'll reach out soon.