TAFFEE: The Agent Factory Framework for Enterprise Enablement
Bootstrap your enterprise AI, without sinking the bootstrapping cost.
An open foundation for enterprise AI agents. Up and running in hours, entirely inside your enterprise. A path to production, not a prototype.
The cost you skip
Building the foundation yourself ⓘOur estimate: five engineers for six months at about $200K per engineer per year, fully loaded (US mean software-developer wage from the BLS, plus benefits).
~$500K · ~6 months
of engineering before the first agent delivers value
With TAFFEE
Up in hours
Start from an open foundation, inside your enterprise. Free and Apache-2.0.
The TAFFEE Opportunity
Why a shared foundation, and why now.
Agents are going mainstream faster than enterprises can build the foundation underneath them. Select a step to see the evidence.
Step 1 · Demand is surging
Demand is surging.
Spending on AI is accelerating, and agents are moving out of pilots into real work. Every team that starts an agent project needs the same groundwork underneath it.
$176.8B
Enterprise spending on using AI
Annualized as of September 2026: 5.4× the $33B spent in 2025.
Source: Menlo Ventures, 2026
62%
of organizations are building, testing or deploying AI agents, up from 53%
Step 2 · The foundation isn't there
The foundation isn't there.
Before agents can be trusted with real work, enterprises need AI-ready data, clear policies and controls. Most do not have them yet, so each team improvises its own.
Step 3 · Projects stall, costs overrun
So projects stall and costs overrun.
Teams spend months on groundwork before an agent delivers value. Many projects never reach production, and budgets overrun along the way.
46%
of AI proofs of concept scrapped before production, at the typical surveyed organization
Step 4 · Every team builds alone
Every team builds alone.
Most AI is bought from many vendors, with nothing shared underneath. Each team wires up its own context, credentials and controls, and the risk adds up.
TAFFEE · The opportunity
Build the foundation once. Every team builds on it.
Before the first agent does useful work, every enterprise builds the same five things. Build them once, as open software inside your enterprise, and every team after the first starts from a working base.
Context
One shared, up-to-date view of your code and knowledge for every agent.
Capabilities
Connect once to your tools, APIs and data, and reuse them everywhere.
Governance
One policy for what agents may do, with credentials and an audit trail.
Orchestrator
Run agents reliably, check their work and stay within budget.
Telemetry
One view of what every agent did, what it cost and how long it took.