From Spec to Production — Zero Handoffs
See what changes when AI agents handle the grunt work.
The same project. Two different realities.
The Old Way
- closeDeveloper reads Jira ticket
- closeOpens IDE, context-switches between files
- closeWrites code for 4 hours
- closeOpens PR, waits 2 days for reviewer
- closeReviewer requests changes
- closeBack-and-forth for 2 more days
- closeFinally merges
The Viloforge Way
- check_circleYou write a spec with acceptance criteria
- check_circleExecutor agent claims and starts immediately
- check_circleClaude Code implements in 8 minutes
- check_circleJudge agent reviews independently
- check_circleRejected? Executor reworks automatically
- check_circleLoop runs until spec is met
- check_circleYou review the final result and merge
3 days, 6 context switches
45 minutes, 1 human decision
Watch the full loop — from spec to done
viloforge-cli
08:00 [YOU] vtf workplan create "Auth System v2" --milestones 3
08:01 [YOU] vtf import phases/auth-v2/ --workplan 12
08:01 [BOARD] 6 tasks imported across 3 milestones
08:02 [BOARD] Milestone 1 activated: "Core Auth"
08:05 [VAFI] executor-01 polling... TASK-047 claimed
08:05 [VAFI] Cloning vilosource/auth-service @ main
08:06 [VAFI] Claude Code working... 47 tool calls, 3 files modified
08:13 [VAFI] pytest: 38 passed, 0 failed
08:14 [VAFI] TASK-047 → pending_completion_review
08:15 [JUDGE] judge-01 claimed TASK-047 for review
08:18 [JUDGE] Independent test run: 38 passed
08:20 [JUDGE] Code review: 12/12 criteria met → APPROVED
08:20 [BOARD] TASK-047 → done ✓
08:22 [ATLAS] executor-02: "what are the auth requirements in the wiki?"
08:22 [ATLAS] → Querying Confluence... 3 results returned
08:23 [VAFI] executor-02 claimed TASK-048
08:35 [VAFI] TASK-048 → pending_completion_review
08:38 [JUDGE] TASK-048: 2/8 criteria failed → CHANGES REQUESTED
08:39 [VAFI] executor-03 reclaimed TASK-048 with feedback
08:47 [VAFI] TASK-048 reworked → pending_completion_review
08:50 [JUDGE] TASK-048: 8/8 criteria met → APPROVED
08:50 [BOARD] TASK-048 → done ✓
12:00 [BOARD] Milestone 1: 4/4 tasks done
12:00 [BOARD] Milestone 2 activated: "Token Management"
12:01 [VAFI] executor-01 polling... TASK-051 claimed
15:30 [BOARD] All milestones complete. 6/6 tasks done.
15:31 [CONFIG] Switching harness: claude-code → codex-cli
15:31 [VAFI] executor-01 reconfigured. Specs and gates unchanged.
15:32 [STATUS] Ready for your review.
Every decision. Every tool call. Replayable.
Built-in execution tracing captures the complete agent session for every task — every prompt, every file edit, every test run. Click any task to replay the full conversation. Debug failures in seconds. Prove compliance to auditors. Know exactly what shipped and why.
Powered by CXDB, an open-source execution trace database by StrongDM.
trace replay — TASK-047
Turn 1/47 — 08:05:12
[system] Read .vafi/context.md and implement the task
Turn 2/47 — 08:05:14
[assistant] Reading the task spec...
[tool] Read src/auth/tokens.py (142 lines)
Turn 15/47 — 08:08:42
[assistant] Implementing refresh token rotation...
[tool] Edit src/auth/tokens.py — add rotate_refresh_token()
Turn 42/47 — 08:12:30
[tool] Bash: pytest src/tests/ — 38 passed, 0 failed
... 47 turns | 12,840 tokens | 8m12s total