Agentic frameworks we build
A skill makes an agent better at one thing. A framework makes a whole recurring job repeatable: the kind of job that currently depends on one person's judgement, takes weeks, and produces something slightly different every time.
We build these as repositories your team clones and runs, with the human deciding at the points that need deciding. The ones below are ours, in daily use, and they descend from one another, which is the point. The second one costs a fraction of the first.
Featured frameworks we have built
Groundwork
The base the others are built on. For teams who want agent work to follow a reviewable shape rather than a chat log.
Bidline
For organisations that bid regularly and cannot spare a senior person for three weeks every time a competition opens.
Pricepoint
For deciding what to charge, when the answer has to survive a board asking where the number came from.
Throughline
For a product whose positioning, book series, website and campaign all have to say the same thing, and keep saying it as the product moves.
Milestone
For a research consortium with deliverables due on a schedule and partners waiting on each one.
What they have in common
Different jobs, same anatomy. This is what we would build for yours.
Named steps, not a chat window
Each stage is a command your team runs deliberately, in order, and can stop at. The work is inspectable between steps rather than emerging whole from one long conversation.
Parallel where it pays
Sub-agents handle the parts that genuinely are independent (one per bid question, one per source, one per market), so a week of reading becomes an afternoon, without them quietly agreeing with each other.
An adversarial stage that must be answered
Built into the workflow rather than left to whoever remembers. A plan is attacked before it is built, and the findings are recorded with a decision against each.
Memory that outlives the session
What was decided and why persists across machines and across weeks, so on Monday the framework knows everything it knew on Friday.
Checks that can fail
Every framework we build carries verification that can actually come back negative, and does. A pipeline whose checks always pass is not checking anything.
Yours to run and change
A repository you own, not a service you rent. Your team reads it, edits it, and runs it without us.
A return you can see
A framework earns its keep on the jobs your team repeats: the bid, the pricing review, the research pack that currently takes a week of someone's best judgement. Carried by agents in named steps, that work can run in an afternoon, comes out the same shape every time, and frees your specialists for the decisions only they can make.
And you will not have to take our word for it. Every framework we build shows its return in your own numbers, from the first run.
A baseline before we build
We time and cost the job as your team does it today, on real recent examples. Every later figure is compared against that, so every improvement shows up as a number you can take to the board.
The same job, run both ways
The framework runs on work your team has already done, and the results are set side by side: hours, cost, and whether the output met the standard you set.
Value defined up front
We agree with you what success is worth before any build starts, so the case for rolling it out further is already made, in evidence you own.
What it costs, and how to start
Tell us what you are trying to get your agents to do. We will come back with whether we can help, what it would involve and what it would cost. No payment at this stage, and if a skill pack would serve you better we will say so.