# Agent Coordinator pitch

> Turn subsidized reasoning capacity into continuous progress on critical work

- HTML version: https://robbiepalmer.me/projects/agent-coordinator/deck
- Project: Agent Coordinator (https://robbiepalmer.me/projects/agent-coordinator.md)
- Slides: 9

## Slide 1

# Put every spare reasoning token on the critical path

Agent Coordinator turns subsidized model capacity into useful parallel work.

## Slide 2

## Reasoning capacity is perishable

Subscription allowances reset. Rate limits recover. Agents become available.

Meanwhile, ready work waits because nobody notices that a critical path has opened.

Unused capacity and delayed work are the same missed opportunity.

## Slide 3

## When a blocker clears, useful work should fan out

```mermaid
flowchart TB
Open[Critical-path blocker clears] --> Research[Research]
Open --> Build[Build]
Open --> Validate[Validate]
Research --> Next[The next constraint moves]
Build --> Next
Validate --> Next
```

Parallelism matters when every branch shortens the route to the outcome.

## Slide 4

## Planning decides what matters. Coordination spends the capacity.

### Work Graph

Priority, dependencies, readiness and the critical path.

### Agent Coordinator

Actor, model tier, reasoning level and available plan allowance.

The plan stays durable. Allocation adapts whenever work or capacity changes.

## Slide 5

## Match the task to the right depth of thought

* Reserve frontier models and high reasoning for ambiguous, consequential work.
* Send bounded implementation to capable mid-tier models.
* Pack routine work into fast, cheap or already-subscribed capacity.
* Route around exhausted plans before buying overflow.

Use the least expensive route likely to finish the task well.

## Slide 6

## Start personal. Scale to teams. Open to contributors.

1. **Personal:** keep one person's agents productive across several subscriptions.
2. **Team:** share capacity under explicit budgets, access rules and concurrency limits.
3. **Contributor:** match public work to people using their own agents and plans.

Each stage adds capacity without giving away control of the plan.

## Slide 7

## The first release is a recommendation engine

It watches ready work, available agents and remaining plan capacity.

When the critical path changes, it proposes the next assignment and explains why.

A person approves the match. Automatic routing comes after the recommendations prove useful.

## Slide 8

## Win by moving valuable work sooner

* Critical-path work spends less time waiting.
* Ready branches start in parallel when suitable capacity exists.
* Less paid-for model capacity expires unused.
* Fewer assignments stall, exceed limits or need reassignment.

The measure is faster progress on outcomes.

## Slide 9

## Keep the critical path supplied with reasoning

As work opens, allocate the best available combination of agent, model, reasoning depth and plan capacity.

Then do it again when the plan changes.

[Read the full project account](/projects/agent-coordinator?tab=overview)

---

Markdown index of this site: https://robbiepalmer.me/llms.txt