Work
How we build and release, and what you can check.
Mode2 is new, so there are no client case studies here yet. What you can judge today is the Foundation every client runs on, how it is released, the terms of our first two deployments, and the method we hold ourselves to.
One foundation under every client
The Mode2 Foundation has four modules. Every offer installs or maintains one. The Code module and the agent base are in their first install.
Code
Managed-settings baseline, hooks and skills, review gates and a test and metrics harness.
Controls
Spend caps by group and member, SCIM, audit log export, the Compliance API, a retention policy and the 12-month bill model.
Agent base
Connectors, human approval steps, audit log, cost controls and evals and a model-swap procedure, on the Claude Agent SDK, running against the Claude API, Amazon Bedrock or Google Vertex.
Governance
AI use policy, model-change procedure, incident runbook, vendor and usage review and board report.
How we release
A release within 30 days of each Anthropic model or major tool change.
- Evaluate the new model against your evals
- Update settings and controls
- Adopt new Anthropic features
- Re-baseline cost
Release notes are published here as each release ships.
Reference-deployment terms
The first two Code Pod clients pay $30,000 in place of the list price. In exchange, they agree to a named public case study with before-and-after metrics.
The metrics are the ones every Pod uses: sessions, lines, commits and PRs from Anthropic's analytics, plus cycle time and defect rate from the client's own Git and tracker.
The method
- 01
A fixed price and a time promise.
Every offer has a published price and a date it ends.
- 02
A named engineer before a build is sold.
You meet the person who builds it before you sign.
- 03
Every agent built so the model can be swapped.
Your agent outlives any one model release.
- 04
Outcomes measured before and after.
We record the numbers first, so the change is visible.
- 05
One foundation under every client.
Code, Controls, Agent base and Governance, maintained by Mode2, so each engagement starts further along.
- 06
A release within 30 days of every Anthropic change.
The new model is tested against your evals, settings and controls are updated, new features are adopted and cost is re-baselined.
Open source
An engineering harness for Claude Code: people own the task specs and acceptance tests, Claude implements each task in its own worktree, and hooks, replay gates and a separate reviewer check the work before a person approves the merge. Retired once Claude Code shipped its own orchestration.
A privacy-first platform for personal AI agents on your own hardware. Retired.
Case studies
Each one goes here when its client signs off on the numbers.
Book a 30-minute scoping call.
Thirty minutes on what you need, which offer fits, and the fixed price for it.