ForthClover
AI Coding Enablement · For engineering teams

Coding agents for the whole team, with guardrails.

Chances are some of your engineers already use Claude Code, Codex, Cursor or Copilot, each with their own settings and no shared rules. In two weeks we roll out the right tool to up to 15 engineers, configured with permissions, sandboxing, repo instructions and CI checks, and train the team on real tickets from your backlog.

See it on your workSee the guardrails
Price
$6,000
Fixed. No day rates.
Timeline
2 weeks
Ten business days
Team size
Up to 15
Then about $300 per engineer
Tools
Neutral
Chosen for your stack
Pick a use case

Where coding agents help, and where your engineers stay in charge.

The workshops cover the use cases your team picks, on real tickets from your backlog. Every session runs inside the same guardrails.

Feature work

Well-scoped tickets taken from plan to draft pull request, following your existing patterns.

The agent does

  • Reads the ticket, the relevant code and your repo instructions
  • Proposes a plan before editing anything
  • Writes the change and its tests on a feature branch

Your engineer does

  • Picks the ticket and approves the plan
  • Reviews the pull request like any other
  • Owns the merge

Example session

Illustrative
  1. > Add a CSV export to the invoices page. Follow the existing reports export.
  2. Read AGENTS.md, the reports export module and its tests
  3. Plan: 5 files, 1 new endpoint. Waiting for approval before editing
  4. Edited 5 files on branch feat/invoice-csv
  5. npm test -- invoices · passing
  6. Opened a draft PR with a summary and test notes

Guardrails in play

Feature branch onlyPlan before editDraft PRs

Signals to watch

  • PR cycle time
  • Review rounds per PR

Human checkpoint

An engineer approves the plan and reviews the PR.

What you get

A configured team, not a licence key and a link to the docs.

Everything is written down and checked into your repos or wiki, so it keeps working after we leave.

1Approved tools & plans

2Permission defaults

3Secrets & data rules

4Review rules for AI code

Day 4

Policy pack

Which tools and plans are approved, what agents may do by default, how secrets and customer data are handled, and how AI-written code gets reviewed. Short enough to actually read.

allow tests, lint, typecheck

ask   installs, network

deny  .env, prod, push main

Day 5

Configured tools

Licences assigned and the agent configured for your team: permission rules, sandbox and network settings, and organisation-wide settings where the tool supports them.

  • Issue tracker · read
  • Internal docs · read
  • Logs & traces · read
  • Design files · read

Day 6

MCP connections

Agents connected to the tools engineers already use, such as the issue tracker, internal docs and observability, through MCP servers with read-only access by default.

secret-scanpassed

testspassed

human approvalrequired

Day 7

Repo conventions and CI guardrails

Agent instruction files (AGENTS.md, CLAUDE.md or the tool's equivalent) in your main repos, plus CI checks: secret scanning, required tests and a human approval on every merge.

Workshop 1 · daily work

Workshop 2 · bigger changes

Days 7 and 9

Workshops, recordings and playbook

Two hands-on workshops on real tickets from your backlog, recorded, plus a playbook of the prompts, patterns and rules that worked for your codebase.

Baseline
Day 30

Day 2 · about 30 days later

Baseline and readout

A baseline of the cycle-time metrics you already track, taken before rollout, and a readout comparing them about 30 days after handover. Exact timing agreed at kickoff.

How it fits together

The agent works through a permission layer, not around it.

Engineers keep their editor and terminal. The agent reaches your repos, CI and tools only through settings your team controls.

  1. 01

    Developers

    In the editor or terminal they already use

    EditorTerminalCloud tasksPull requests

  2. 02

    Coding agent

    One tool, or a mix, chosen with you

    Claude CodeOpenAI CodexCursorGitHub Copilot

  3. 03

    Permissions & sandbox

    Enforced, not just suggested

    Org-wide settingsCommand allowlistsFilesystem sandboxNetwork rulesHooks

  4. 04

    Your repos, CI & tools

    Reached only through the layer above

    Git repositoriesCI pipelinesMCP serversPackage registries

Across every layer

Audit logs

Agent activity and PR provenance recorded where the tool supports it.

Across every layer

Secret scanning

Pre-commit and CI scans; secrets kept out of agent context.

The two weeks

Configure first, then teach on real work.

Ten business days. A small pilot group tries the setup mid-week one, so the workshops run on a configuration that already works.
  1. D1–2Baseline & tool choiceCurrent metrics, repo review, options compared
  2. D2–5Policies & configurationPermissions, sandbox, secrets, licences
  3. D4–7MCP & repo conventionsTool connections, agent instruction files
  4. D5–7CI & review guardrailsSecret scanning, required checks, approvals
  5. D6–9WorkshopsTwo hands-on sessions on your backlog
  6. D9–10Playbook & handoverPlaybook, recordings, readout plan
Guardrails

What agents may do on day one.

Our starting defaults. We tune them with your security lead and enforce them in the tool's settings and in CI wherever possible, because instruction files alone are guidance, not enforcement.

Allowed by default

  • Read code in the repository it was opened in
  • Edit files on a feature branch
  • Run tests, linters and type checks
  • Open draft pull requests

Asks first

  • Install or upgrade dependencies
  • Run commands outside the allowlist
  • Reach the network beyond approved domains
  • Use write actions on connected tools, such as updating a ticket

Blocked by default

  • Push to main or merge its own pull requests
  • Read .env files, keys or secrets stores
  • Touch production databases or credentials
  • Skip or disable CI checks, hooks or tests
Who does what

What we need from you, and what we take off your plate.

You provide

  • An engineering lead as owner, about three hours a week
  • Admin access to your Git host and CI, or someone who has it
  • Licences bought on the plan we agree
  • The cycle-time metrics you track, or access to the tool that tracks them
  • Engineers' time for two workshops

We provide

  • A senior engineer from tool choice to handover
  • Policy pack and tool configuration
  • MCP connections and agent instruction files
  • CI and code-review guardrails, secret scanning
  • Two workshops, recordings, playbook and the readout

Not included

  • Licence and model usage fees (billed by the vendor)
  • Engineers beyond 15 (priced per engineer)
  • Custom MCP servers for in-house systems (quoted)
  • Ongoing administration after handover (see AI Operations)
Is this you?

Built for engineering teams that want speed without surprises.

A good fit if…

  • You have five or more engineers working in Git with CI
  • Some engineers already use AI coding tools, without shared rules
  • Security or leadership wants guardrails before a wider rollout
  • You track, or want to track, metrics like PR cycle time

Not the right start if…

  • You have fewer than about five engineers; a self-serve setup is usually enough
  • There's no CI or code review process yet
  • Policy forbids sending code to any third-party AI service
  • You want a guaranteed productivity number before you start
Where it sits

From faster engineers to an agent-ready product.

Once your team works well with agents, make your own product usable by them, test what you ship, and keep it running. How we build every system
  1. Step 1 · you are hereAI Coding Enablement$6,000fixed · 2 weeks · up to 15 engineers
  2. Step 2Agent-Ready IntegrationFrom $6,500fixed · 2–3 weeksDetails
  3. Step 3Agent Security Review$3,500fixed · 1 week · one agentDetails
  4. Step 4AI OperationsFrom $990/mothree tiers · 3-month minimumDetails
FAQ

Questions engineering leaders ask us

Which tool do you recommend?
It depends on your stack, your editors and your security requirements. We're tool-neutral: in the first days we compare options such as Claude Code, OpenAI Codex, Cursor and GitHub Copilot against your repos and policies, and recommend one, or a mix if different teams work differently.
Will our code be used to train AI models?
We set you up on business or enterprise plans that exclude your code from model training where the vendor offers that, and we walk you through each vendor's current data-use terms before you buy. Terms differ by vendor and plan and they change, so your legal or security lead should confirm them.
How do we measure impact?
With metrics your team already tracks, such as PR cycle time, time to first review or deployment frequency. We record a baseline before rollout and compare about 30 days after handover. We report what moved and what didn't. We don't promise a number.
Is it secure?
Agents run with least privilege: sandboxed where the tool supports it, with command allowlists and no access to secrets or production. Changes reach main only through your normal review and CI. We put the rules in enforced settings and CI checks, not only in instruction files, and add secret scanning.
Do licences cost extra?
Yes. Licences and any usage-based model fees are billed by the vendor directly to you. We help you pick the plan and the seat count. The $6,000 covers our work only.
What if we have more than 15 engineers?
Each additional engineer is about $300, covering setup and a workshop seat. Larger teams usually run the workshops in more than one group.
$6,000 · fixed · 2 weeks · up to 15 engineers

If your engineers already use coding agents, give them shared rules.

On a free fit call we'll look at your stack, your current tool use and your security requirements, and tell you what a two-week rollout would cover.

Book the fit call