Agent Deployment Kit
Roll out always-on AI agents without losing control.
An always-on agent acts inside your systems while nobody is watching. The controls that make that safe are familiar: identity, access, approvals, audit and evaluation. This kit gives you working versions of each, ready to adapt for your first agent pilot.
Coming soon. Ask to be notified at launch.
What is inside
7 documents, about 3,900 words, each as DOCX, Markdown and PDF.
00Start here: the rollout in six steps
Which document to use at each step, from choosing the job to choosing the platform.
~400 words
01Approval policy as code (Python)
Four risk tiers in code with a test suite security can sign off; a model reviewer can escalate but never downgrade.
~800 words
02Agent identity checklist
18 checks for a separate identity, scoped short-lived credentials, isolation, audit and revocation.
~550 words
0390-day agent pilot plan
Set-up, shadow, supervised and review stages with exit criteria, roles and stop conditions.
~550 words
04Audit log schema
Event fields and a JSON example your SIEM can ingest, plus alert rules and useful queries.
~550 words
05Agent evaluation scorecard
Accuracy, policy adherence and injection resistance targets before widening permissions.
~400 words
06Vendor evaluation worksheet
Weighted criteria for comparing hosted agent products and your own build, and questions for every vendor.
~600 words
Who it is for
- Platform and IT teams preparing to run agents in production
- FDEs and consultants deploying agents for clients
- Security and risk teams approving an agent rollout
Who it is not for
- You want a ready-made agent product; this is governance and rollout material
- You are building a chatbot that takes no actions
- You need legal or compliance advice for your jurisdiction
Agent Deployment Kit FAQ
Which agent products does it cover?
It is vendor-neutral. The vendor worksheet compares any hosted agent product (such as OpenAI Dots, Meta Muse, xAI Grok Bot, Google Gemini Spark, Microsoft Copilot Autopilot or Claude) and a custom build on isolation, identity, approvals, audit and data handling.
Is the code ready to use?
The approval policy is a working Python starter with a test suite. Adapt the tool lists and limits to your agents and run the tests in your CI.
I already own the FDE Toolkit. Is this the same?
No. The FDE Toolkit includes three one-page, customer-facing agent templates for engagements. This kit is the full implementation: policy code and tests, the identity checklist, audit schema, evaluation scorecard and vendor worksheet.
Launching soon
Ask to be notified and you will hear the day it launches, at the price shown.