Confidential

Version control for agents.

The new reality

Agents write a growing share of our software.

Share of new code written by agents

2023

~10%

Today

~40%

Ahead

~90%

Software is increasingly written by agents, and the shift is only accelerating.

The new reality

Agents now work in parallel and make trying ideas cheap.

agents

Editing

implementing JWT refresh feat/auth

In review

indexing search results feat/search

Done

rate-limiter token bucket feat/ratelimit

icon component refresh fix/icons

test fixtures cleanup chore/tests

oauth provider docs docs/oauth

Idle

JWT spike research spike/jwt

Every agent runs in its own session and branch, shared with the right people. Trying an idea is now cheap.

01   Problem

Git was built for one developer, a few commits a day, alone on a branch.

It’s 2026, and it is still a pain to:

× Parallel work

Run parallel agents without them trampling each other’s work.

× Local context

Share and collaborate on local agent files and context.

× Trust

Trust agents not to delete or break files as they go.

× Tool sprawl

Copy-paste between Notion, Linear, and GitHub just to collaborate.

02   Problem

Git versions a sliver. Everything else agents touch is invisible.

Versioned today

Code

Not versioned at all

Environment Plans Decisions Context Agent work Reviews & comments

Git was built for one human, working sequentially. So teams throttle their agents instead of unleashing them. The bottleneck is no longer the model, it’s the infrastructure to run, review, and trust what agents produce.

03   Current solutions

Today, teams duct-tape four tools. It still breaks.

× Git + GitHub: versions code and nothing else. Built for one human committing in sequence.

× Notion: where plans and specs live. Not versioned, not connected to the code they describe.

× Linear: where work gets tracked. Disconnected from the actual files agents are changing.

× Agent IDEs (Cursor, Devin): run agents, but no shared workspace, no team review, no memory across runs.

Four disconnected tools, none built for agents. The work agents actually do falls through the gaps between them.

04   Vision

Self-improving agents.

The next leap isn’t a smarter model. It’s agents that upgrade themselves. To do that safely, an agent has to branch off, change its own code and environment, and test the upgrade in production before keeping it.

Branch the live system → upgrade itself → test in production → keep what works

None of this is possible on today’s tools. Self-improving agents need far more powerful version control: branch, run, and merge the whole system, not just code. That’s the infrastructure Mod is building.

05   Solution

Mod is version control for the whole system.

Connect a local agent → session auto-created → tracked in parallel → structured context

06   Demo

Your collaborative agent harness.

Sessions Canvases Connect your agents Checkpoints & revert Extensible harness

07   Market

The market expands as software starts being grown.

$1T+

$300B

$80B

$80B · Now

Agent infrastructure for engineering teams. The workspace teams build with agents in.

$300B · Next

All software development. As agents do the work, every team’s tooling spend follows them.

$1T+ · Expansion

Every business runs its own software. When software is grown, not shipped, the buyer is no longer just engineering.

Agent compute is already a line item and spend per team is compounding fast. The next buyer isn’t a developer at all.

08   Competition

Productivity apps bolt AI on.
Mod is the infrastructure underneath.

The market we take

Productivity & collaboration apps, built for the human era.

What the agentic age needs

Productivity apps

Mod

Versions the whole system, not a silo

◐

●

Agents as first-class collaborators

◐

●

Parallel agent work: branch & merge

✕

●

Checkpoint, attribute & roll back

✕

●

Local-first, runs on your infrastructure

✕

●

One programmable substrate, not a fixed app

✕

●

An app can add AI features. It can’t become infrastructure. That gap is the whole opportunity.

09   Team

Built by the people who built the primitives.

Nassar Hayat

Co-founder & CEO

Head of Platform at MultiOn, building browser agents. Led the developer ecosystem team at Radicle, a peer-to-peer GitHub alternative. Co-founded Foodchain, a B2B platform bootstrapped to $5M ARR.

Alex Good

Co-founder & CTO

Lead maintainer of Automerge, the leading open-source CRDT library, at Ink & Switch. Led the protocol team at Radicle. Co-founded Foodchain. Speaker at Local-First Conf and QCon London.

Years of R&D on local-first collaboration and agent architectures. We have shipped together before. Longtime collaborators, and each other’s best man.

10   Business model

We sell the infrastructure
agents run on.

Revenue (ARR) Teams

We launch in ~3 months. Year-3 target: Cursor’s ~50k team scale and $250M ARR (~$5k per team).

Free Get started

Collaborative workspace

Real-time sync

Backup storage

Bring your own model keys

Pro For teams

Everything in Free

Bundled inference

Storage and hosted VMs

3× model usage

Max Agents at scale

Everything in Pro

20× inference

Dedicated VMs

Priority sync

Every plan bundles OS inference Storage Hosted VMs Sync

11   Traction

Live today. Technical leaders are already pulling for it.

“Give me a way to share my agent data with the team and I will stop using Linear, GitHub, and Notion.”

Engineering leader

“Our product and engineering teams are stuck in so much grunt work trying to organise and collaborate on plans.”

Engineering leader

“We’re struggling to manage agents across Notion for specs, Linear to track, and a separate GitHub repo for local context.”

Engineering leader

Quote placeholder

Name, role

Quote placeholder

Name, role

Quote placeholder

Name, role

Quote placeholder

Name, role

Quote placeholder

Name, role

Quote placeholder

Name, role

Shipped and in production: the CLI (npm install -g @mod/mod-cli) and the real-time collaborative workspace.

12   The ask

Raising to reach the first 1,000 teams.

$X.XM

~18 months of runway to grow the team below and put Mod in front of its first 1,000 teams.

Use of funds

Product Engineering2 hires

Infrastructure Engineering2 hires

ML Engineering2 hires

Infrastructure & inferencecompute

Sales & marketingGTM

mod.computer nas@mod.computer