Who leads
Task routing
Route the work to the platform most likely to succeed first. A CLI-first code edit task might start in Aider or Codex, while cross-surface rule cleanup might start in Claude or Cursor.
Experimental— Research-phase static routing rules. Harmony is the GA cross-platform operator surface; Synergy is not at the same maturity yet.
Synergy — Experimental
1 + 1 + 1 should be greater than 3.
Synergy is the Nerviq model for multi-agent lift. It explores which platform should lead, which supporting platforms should contribute evidence, and how one tool can compensate for another tool’s blind spots. It is still an experimental layer, not the same production-grade surface as Harmony.
ExperimentalResearch phaseTask routingCompound evidence
Start with Harmony first
Harmony is the stable operator surface for cross-platform governance, drift detection, and sync decisions. Start there for production workflows. Synergy is best used after that baseline is already working, when you want to explore routing and multi-agent lift hypotheses. Read Harmony first.
What Synergy Means
A repo with multiple agents is not automatically better. Synergy turns a stack of tools into a system with explicit task routing and evidence accumulation, but the model is still experimental and should be treated as advisory.
Lead
Primary platform
One agent gets the first shot at the task based on current strengths and context fit.
Support
Companion platforms
Additional agents contribute checks, evidence, or verification where they are strongest.
Score
Measured gain
Synergy compares single-platform and compound-platform evidence so teams can see when the extra complexity is worth it.
Core Concepts
Synergy is built around three ideas: route smart, accumulate proof, and let one platform offset another platform’s weak spots.
Who confirms
Compound evidence
Nerviq records more than “did the task pass.” It combines findings from multiple platforms so the final recommendation carries stronger proof than any one surface alone.
Who covers whom
Compensation matrix
Every platform has blind spots. Synergy makes those blind spots explicit and pairs them with platforms that compensate for them.
Compensation Matrix
This is the practical heart of Synergy: some platforms lead, others strengthen the call.
| Concern | Primary platform | Companion platform | Why the pair is stronger |
|---|---|---|---|
| Security posture | Codex | Claude | Codex catches trust boundaries; Claude reinforces repo-local rules and hooks. |
| Realtime editor drift | Cursor | Claude | Cursor catches rule-level issues while Claude checks repo-owned instructions and command loops. |
| Headless automation | Aider | OpenCode | Aider is strong for CI edits; OpenCode adds command orchestration and server-mode context. |
| Enterprise governance | Copilot | Windsurf | Copilot contributes policy/org signals, while Windsurf adds stronger enterprise segmentation and memory caveats. |
The synergy-report Command
Use Synergy when you want one report that answers which platforms are active, what each one is good at, and where combining them might produce a better outcome. The key word is might: this surface is still experimental.
bash
nerviq synergy-report --format markdown
# Example summary
Active platforms: claude, codex, cursor
Lead platform for trust and automation: codex
Lead platform for instruction quality: claude
Lead platform for rule hygiene: cursor
Combined gain:
- 3 platforms surfaced 11 unique findings
- Best single platform surfaced 7 findings
- Net lift: +57%Example: Three Platforms Beat One
Here is the kind of situation Synergy is built for: the same repo uses Claude, Codex, and Cursor. Read this as a research model for multi-agent lift, not as a guaranteed production routing answer.
Single-agent view
What each one catches
- Claude catches missing verification commands and repo-owned instruction drift.
- Codex catches approval, sandbox, and trust-boundary mismatches.
- Cursor catches MDC rule failures, privacy-mode gaps, and editor-specific MCP mistakes.
Synergy view
What the combined view adds
- The same verification policy is now visible across repo, CLI, and IDE surfaces.
- Contradictions are surfaced instead of hidden because each platform sees the repo through a different boundary.
- The final recommendation carries more confidence because three evidence streams point at the same fix.