Execution Intelligence
Traditional planning is one-directional: create a plan, then execute. The plan never learns from execution. When estimates are wrong — and they always are — the schedule silently drifts.
Execution intelligence closes this loop. Every signal from development feeds back into Plenar, and the plan recomputes.
The feedback loop
Plan → Execute → Signal → Recompute → Plan
↑ ↓
Calibrate ← ─ ─ Actual vs Estimated- Plenar computes your plan from constraints
- You execute — start tasks, complete tasks, hit blockers
- Each action is a signal that updates the plan
- The plan recomputes with real data
- Over time, estimate accuracy improves
Signals that feed back
Every interaction updates the plan:
| Signal | What happens | How it enters |
|---|---|---|
| Task started | Start date is anchored. Downstream tasks lock to this date. | plenar_start_task or dashboard |
| Task completed | Actual duration recorded. Dependent tasks can begin. Schedule recomputes. | plenar_complete_task or dashboard |
| Blocker reported | Task marked blocked. Plenar surfaces alternative tasks to work on. | plenar_report_blocker or dashboard |
| Estimate revised | Timeline shifts for the task and everything downstream. | plenar_update_estimate or dashboard |
| PR linked | Code connected to task. File paths indexed for exemplar matching. | plenar_link_artifact |
| Time off added | Capacity reduced for affected days. Tasks shift. | Calendar sync or plenar_add_time_off |
| Scope changed | Tasks added or removed. Full recompute. | plenar_add_tasks, plenar_delete_tasks |
None of these require manual schedule updates. The plan stays current because signal flows are part of the natural development workflow.
Estimation calibration
Plenar tracks two values for every completed task:
- Estimated days — what you predicted at planning time
- Actual days — working days between effective start and effective end
This builds a calibration history per team. Over time, patterns emerge:
- Tasks consistently take 1.4x the estimate → your team underestimates by 40%
- Backend tasks are accurate, frontend tasks take 2x → different calibration by type
- A specific team member’s estimates are consistently tight → per-person calibration
Velocity context
Use plenar_get_velocity_context to see your team’s calibration data:
- Calibration factor — ratio of actual to estimated across completed tasks
- Accuracy trend — whether estimates are improving or degrading
- Throughput — tasks completed per week, working days consumed
This data helps you make better estimates for new work. If your calibration factor is 1.4, a task you’d estimate at 3 days should probably be estimated at 4-5 days.
Exemplar matching
Use plenar_get_exemplars to find completed tasks similar to one you’re about to estimate:
- File path overlap — if you’re estimating a task that touches
src/api/auth/, Plenar finds past tasks that touched the same files - Task type matching — filter by type (e.g., “endpoint”, “migration”, “UI component”)
Each exemplar shows: title, estimate, actual duration, and file paths. This gives you concrete reference points instead of guessing.
Agent participation
When connected via MCP, Claude Code agents generate signals automatically as part of their workflow:
- Starting work →
plenar_start_task→ plan knows work began - Finishing →
plenar_complete_task→ actual duration captured, next task suggested - Stuck →
plenar_report_blocker→ alternatives surfaced - Shipping →
plenar_link_artifact→ PR linked, file paths indexed for future exemplars
The agent doesn’t maintain the plan — the plan maintains itself from the agent’s natural workflow signals.
Integration signals
Beyond agent interactions, external systems also feed into the plan:
- Google Calendar — time off and events sync automatically. If someone blocks off a week, Plenar adjusts their capacity and shifts their tasks.
- Jira — status changes sync bidirectionally. Completing a task in Jira updates the plan.
- SCIM — team membership changes from your identity provider flow into project assignments.
See Integrations for setup.
Why this matters
Without execution intelligence, a plan’s accuracy degrades from the moment it’s created. Every day of execution introduces drift that nobody corrects.
With it:
- The plan reflects reality, not the original assumptions
- Estimates improve because you have calibration data and exemplars
- No manual schedule maintenance — signals do the work
- Stakeholders see a live plan, not a stale snapshot from last month’s planning session
The goal: your plan should be more accurate on day 30 than it was on day 1. Execution intelligence makes that possible.