A clearer, faster way to work with OpenCode on real software projects. Pantheon brings planning, implementation, review, and documentation into one guided, token-efficient experience. It is an OpenCode plugin and installer for developers and teams who want useful structure without giving up speed or control of their code.
Português (Brasil) · Repository · MIT License
Pantheon is a companion for OpenCode that helps you move from an idea to a reviewed, verified change. Fourteen specialized agents share one streamlined workflow: plan, implement, check, review, and record what mattered. You stay in charge of every decision and every merge.
flowchart LR
U["You"] --> Z["Zeus<br/>orchestrator"]
Z -->|"delegates"| S["14 specialists<br/>hermes · aphrodite · demeter · themis …"]
Z -->|"ambiguous route"| J["Jev decision router<br/>typed · budgeted · reversible"]
J -->|"orders · ranks · verifies"| S
S -->|"last resort"| M["Native MCP<br/>memory · persistence · vision<br/>code-mode · resources"]
Z --> G["J1 risk gate<br/>off / shadow / warn / enforce"]
G -->|"high risk"| H["Human approval"]
The escalation ladder every agent follows: session context first → free local selector → one typed Jev decision when genuinely ambiguous → native MCP call only as the last resort for state the session cannot reconstruct.
flowchart LR
R["rough request"] --> T0{"clear &<br/>bounded?"}
T0 -->|yes| X["execute directly"]
T0 -->|"vague /<br/>multi-concern"| T1["Tier 1 — internal<br/>minimal adjust ≤15 lines<br/>(silent, no approval)"]
T1 --> T2q{">> given?"}
T2q -->|yes| T2["Tier 2 — full contract<br/>6 sections, confirm-or-adjust"]
T2q -->|no| X
Skill discovery runs a trust cascade — canonical first, popularity never compensates for provenance:
flowchart LR
L["installed pool<br/>(always first)"] -->|"gap"| C1["anthropics/skills<br/>canonical"]
C1 -->|"no fit"| C2["skills.sh<br/>install-count signal"]
C2 -->|"no fit"| C3["GitHub raw<br/>widest, weakest"]
C3 --> J["Jev orders<br/>the candidates"]
J --> P["YOU review<br/>and run the install"]
- Fast-Track execution: Bounded micro-edits (<= 2 files, <= 50 lines) bypass all heavy ceremony and verify in < 10 seconds.
- Mandatory test slicing: Sub-second targeted test execution instead of running sluggish repository-wide test suites during dev loops.
- Assertive council (
/pantheon): Parallel specialist consultations resolved decisively via domain authority and JEV multi-criteria decisions without endless debate turns. - You stay in charge: Pantheon supports your judgment; it never silently alters permissions or runs unapproved destructive actions.
- Non-destructive upgrades: Updating Pantheon (
pantheon-opencode update) refreshes skills, MCPs, and instructions while strictly preserving all user models, presets, permissions, and custom configs.
Requirements: OpenCode 2.x and Node.js >=26.4.0. (See Compatibility).
Option A — Quick Plugin Add (Global or Project):
opencode plugin add pantheon-opencode
# Or pin a specific version:
opencode plugin add pantheon-opencode@2.0.0-rc.1Option B — Global CLI Install with Full MCP Tooling:
npm install -g pantheon-opencode@beta
pantheon-opencode init --yespantheon-opencode doctorRuns a 28-point diagnostic checking MCP initialization, Python virtual environments, Node version, permissions, and config integrity.
When a new version is released:
pantheon-opencode updateThe updater verifies package identity, pulls the latest release, and runs a non-destructive configuration refresh. Your custom models, active preset, custom permissions, plugin options, and user-defined AGENTS.md sections are preserved verbatim.
Pantheon balances speed and rigor through three distinct execution modes:
flowchart TD
Req["User Request"] --> Size{"Scope / Risk"}
Size -->|"≤ 2 files · ≤ 50 lines<br/>routine fix / typo / CSS"| Fast["Fast-Track<br/>(Sub-10s · Inline Test Slicing)"]
Size -->|"Architecture / Tech choice<br/>Security / Multi-stakeholder"| Council["Council Synthesis (/pantheon)<br/>(2-3 Specialists in Parallel · Domain Authority)"]
Size -->|"Multi-phase feature<br/>Heavy refactor · Breaking changes"| Deep["Deepwork (/pantheon-deepwork)<br/>(Athena Plan · Phase Gates · Checkpoints)"]
For routine bug fixes, style adjustments, or localized refactors:
- No planning ceremony: Directly executes bounded changes.
- Targeted test slicing: Runs ONLY the test slice relevant to the modified file (e.g.
node --test tests/foo.test.mjsorpytest -k <func>). Full repository test suites are reserved for CI and release gates. - Single-turn delivery: Verifies and completes in under 10 seconds.
When confronting architectural crossroads, technology evaluations, or security tradeoffs:
/pantheon Should we use WebSockets or Server-Sent Events for live push notifications?
- Parallel Dispatch: Dispatches 2–3 specialist agents simultaneously with independent prompts.
- Assertive Arbitration: Rather than looping in repetitive debate rounds, divergences are resolved immediately via Domain Authority Precedence:
- Security / Auth:
@themishas veto power over vulnerable designs. - Database / Schema:
@demetercarries authoritative weight on query performance and transaction safety. - Frontend / UX / A11y:
@aphroditegoverns user interface responsiveness and accessibility. - Infrastructure / CI:
@prometheusgoverns deployment contracts and container isolation. - API / Protocols:
@hermesgoverns contract consistency and transport schemas. - Architecture / Modularity:
@athenagoverns separation of concerns.
- Security / Auth:
- JEV Multi-Criteria Decision: If trade-offs balance competing requirements equally, Zeus invokes
pantheon_jev_decide(or local decision matrix) to rank candidates deterministically in one tool call.
For multi-phase tasks spanning 10+ turns or multiple sessions:
/pantheon-deepwork "Implement OAuth2 PKCE authentication flow"
- Phase 0 (Plan): Athena maps affected modules and returns a complete Markdown plan displayed directly in the main conversation.
- Phase 1 (Discovery): Apollo performs structured exploration.
- Phase 2–N (Implementation): Specialists implement features in parallel phases.
- Phase Gates: Themis audits each phase (
PASSorFAIL). - Resumability: Checkpoints persisted to
.pantheon/deepwork/<slug>/. Resume anytime with/pantheon-deepwork --resume <slug>.
Every root dispatch carries a concise handoff brief — objective, bounded scope, relevant inputs, constraints, expected return format, and verification. Work is routed qualitatively — uncertainty, domain expertise, dependencies, risk, and verification — not by fixed file or line thresholds. Return contracts are role-specific: implementation uses ## subtask_summary, investigation returns structured findings, Athena returns the complete plan, and council responses use ## specialist_response.
| Agent | Domain Authority | Core Responsibilities |
|---|---|---|
| @zeus | Central Orchestrator | Directs workflow, manages delegations, arbitrates council decisions, enforces token efficiency. Never implements directly. |
| @athena | Architecture & Planning | Research-first planning, system boundaries, LTS policy, dependency analysis. Plan-only; never implements. |
| @hermes | Backend & APIs | FastAPI, Python async, Node/TypeScript backends, API contracts, protocol conformance. |
| @aphrodite | Frontend & UX | React 19, strict TypeScript, responsive design, WCAG accessibility, CSS optimization. |
| @demeter | Database & Persistence | Schema migrations, query optimization, N+1 query prevention, transaction integrity. |
| @themis | Security & Quality Gate | OWASP audits, static analysis (Biome/ruff), authorization gates, dead code removal. Reviews material changes. |
| @prometheus | DevOps & Infrastructure | Docker, CI/CD pipelines, runtime configuration, provider abstraction, telemetry infra. |
| @talos | Hotfix Fast Lane | Rapid, standalone fixes for small bugs, typos, and minor logic without subagent overhead. |
| @apollo | Investigation Scout | Read-only discovery: parallel codebase and web searches. Returns structured findings. |
| @hephaestus | AI Pipelines & RAG | Embeddings, vector stores, prompt engineering, agent pipelines. |
| @gaia | Geospatial & Sensing | Remote sensing, SAR, geospatial analysis, ML/DL raster classification (read-only). |
| @iris | Git & Release Ops | Branching, PRs, issue triage, release notes. Never pushes or merges without human approval. |
| @mnemosyne | Project Memory Bank | Maintains .pantheon/memory-bank/ and records ADRs when explicitly requested. |
| @nyx | Observability & Telemetry | OpenTelemetry tracing, token/cost monitoring, execution profiling. |
- In-Process Skill Search (
pantheon_skill_search): Quickly discovers local skills and external ecosystem skills without requiring shell execution permissions. - 25 Production Skills: Includes
test-slicing,fast-path-execution,tdd-with-agents,instruction-forge, andincremental-implementation. - 5 Local MCP Servers:
pantheon-memory: Shared project memory with atomic SQLite locking.pantheon-persistence: File and context persistence.pantheon-resources: Local documentation and asset serving.pantheon-code-mode: Sandboxed Python execution.pantheon-vision: Image analysis and diagnostic visual inspection.
- Advisory JEV Decision Router: Evaluates candidate options (
pantheon_jev_decide), orders trade-offs, and provides formal multi-criteria decision making without modifying permissions.
Run /pantheon-settings (or Pantheon: Settings in the command palette) to open the interactive settings window:
- Gates & Overrides: Toggle named behaviors live; persists to
pantheon-overrides.json. - Model Presets: Switch active model tier globally or per project.
- Decision Model: Configure local and remote JEV decision routing.
- Updates: Check registry channel (
beta/stable) and trigger upgrades. - Doctor: Run inline system health diagnostics.
- TUI Personalization: Adjust spacing (
comfortablevscompact), separators, gauge bar styles, and context display rows. - Non-Destructive Overlay & Mouse Navigation: Opens as a floating dialog overlay over the active session without blanking the workspace (configurable via
dialogvsfullscreen), with full mouse click support for tabs, sidebar, rows, value cyclers, and actions. - Modernized What's New Window: Card layout with rounded border, pagination buttons, and a dedicated
[OK, Entendido]button with keyboard (Enter/Esc/Q) and mouse click support.
| Start Here | Architecture & Deep Dives | Reference & Operations |
|---|---|---|
| Quick Start | Architecture | MCP Tools |
| Installation Guide | Delegation & Gates | Native Decisions (Jev) |
| Compatibility | Memory System | Model Routing |
| Upgrading | OpenCode 2.x Integration | Changelog |
- Release Candidate Series Inauguration: Production-ready multi-agent orchestration for OpenCode 2.x.
- Agent Token Compaction (-44.3%): Removed duplicated backend code in Apollo/Prometheus, saving 8,150 tokens per turn (~29.3M tokens/month).
- Pragmatism & Anti-Pedantry Directive: Safe diagnostic operations are authorized without bureaucratic gates, ending agent sluggishness.
- Batched Suite TDD & Test-Slicing: Cohesive multi-endpoint testing and sub-100ms targeted test execution.
- Proactive Skill Discovery & Query Caching: Cached skill lookups (24h TTL) and automated recommendation for external tools.
- Live JEV Scorecard & Sandbox Web App: 16/16 pass rate ($0.0000238/decision) verified with an end-to-end sandbox application.
See CHANGELOG.md for full historical release notes.
Ideas, bug reports, and contributions are welcome! Please read CONTRIBUTING.md before submitting a PR.
Pantheon is open-source software licensed under the MIT License. Cite via Zenodo concept DOI: 10.5281/zenodo.22650136. Citation metadata is available in CITATION.cff.
Canonical repository: https://github.com/ils15/pantheon-opencode