testing MCP Server
Architect-centric agentic swarm plugin for OpenCode. Hub-and-spoke orchestration with SME consultation, code generation, and QA review.
Discovered via github-topic:mcp and last synced 3mo ago.
1. Install the package
npx opencode-swarm
2. Add to claude_desktop_config.json
{
"mcpServers": {
"opencode-swarm": {
"command": "npx",
"args": [
"opencode-swarm"
]
}
}
}Config file location: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) / %APPDATA%\Claude\claude_desktop_config.json (Windows)
Unattended multi-interaction runs
Effect
minimax-coding-plan/MiniMax-M2.5
SME agents provide domain guidance
Phase metrics, lessons learned, error taxonomy classification (injected into next phase)
Medium
`minimax-coding-plan/<model>`
Fast
30 min
What It Means
`google/<model>`
Description
Format
reviewer, critic, critic_sounding_board, critic_drift_verifier
Tasks are implemented one at a time through the QA pipeline
Phase-close drift gate: `critic_drift_verifier` verdict (APPROVED/NEEDS_REVISION) and summary (written by architect via `write_drift_evidence`; required before `phase_complete`)
Gate check blocked task completion
`false`
`{ default: 128000 }`
C A -->
New agent session created
Context budget crossed warning/critical threshold
T T -->
A A -->
Default Limit
Legend: Core = always available, Optional = available by default (can be disabled), Conditional = requires specific feature config (ui_review or council) Run `/swarm status` and `/swarm agents` to see what's active. ```mermaid graph TB subgraph Orchestration A[Architect] end subgraph Discovery E[Explorer] end subgraph Execution C[Coder] T[Test Engineer] D[Designer] end subgraph Quality R[Reviewer] CR[Critic] CO[Critic Oversight] CSB[Critic Sounding Board] CDV[Critic Drift Verifier] CHV[Critic Hallucination Verifier] end subgraph Support S[SME] DOC[Docs] CI[Curator Init] CP[Curator Phase] end subgraph Council CM[Council Member] CMO[Council Moderator] end A -->
Existing `.swarm/` state was found, so Swarm continues where it left off
Verdict, risk level, specific issues
Session ends (reason: normal, timeout, error)
Maximum safety — adds slop-detector and incremental-verify hooks
CM CM -->
30-second throttled keep-alive signal
`.swarm/evidence/`, `.swarm/outputs/`
string[]
Runs your project's native build/typecheck
`plan.json`
R R -->
google/gemini-2.5-flash
architect
Architect writes or updates the phased plan (includes CODEBASE REALITY CHECK on brownfield projects)
Secret scan results: findings count, files scanned, skipped files (v6.33)
Task workflow state transitions
QA gate skipped without valid reason
Can Write
string[]
`100`
Offline security analysis, 63+ rules, 9 languages
Architect-only: writes mutation gate results atomically to `.swarm/evidence/{phase}/mutation-gate.json`; accepts verdict (PASS/WARN/FAIL/SKIP), kill rate metrics, and optional survived mutant details; normalizes uppercase-to-lowercase before persisting
A A -->
minimax-coding-plan/MiniMax-M2.5
`kimi-for-coding/<model>`
Swarm asks for missing information it cannot infer
Pass/fail counts, coverage %, failure messages
Agent identity confirmed via chat.message
Coder revision limit exceeded
Skips compaction service — for short sessions under context pressure
Swarm
`anthropic/<model>`
sme, docs, designer
A phase closes out, including: explorer rescan, docs update, `context.md` update, `write_retro`, evidence check, `sbom_generate`, **`@critic_drift_verifier` delegation** (drift check — blocking gate), `write_drift_evidence` call with verdict, mandatory gate evidence verification (`completion-verify.json` + `drift-verifier.json` both required), then `phase_complete`
When Emitted
Phase completed and new phase started
Behavior
Everything
Agent is stopped
Sub-agent returns (success, rejected, error)
Architect wrote outside declared scope
boolean
Catches TODOs, FIXMEs, stubs, placeholder text
Architect-only: generates LLM-based mutation patches (5–10 per function across 6 types: off-by-one, null substitution, operator swap, guard removal, branch swap, side-effect deletion) for direct consumption by the mutation_test tool; returns SKIP verdict on LLM failure rather than throwing
explorer, coder, test_engineer
Critic reviews the plan before execution
Deterministic gate: verifies plan task identifiers exist in source files (written automatically by `completion-verify` tool; required before `phase_complete`)
Evidence written to `.swarm/evidence/{taskId}.json`
Transient error triggered model fallback
Everything (except plan files)
string[]
CycloneDX dependency tracking, 8 ecosystems
Tool
Explorer scans the codebase; co-change dark matter analysis runs automatically to detect hidden file couplings (v6.41)
Files changed, additions/deletions
Task dispatched to a sub-agent
Repetitive tool call pattern detected
Agent is warned, then stopped
Type
Tree-sitter validation across 12 languages
Applies LLM-generated mutation patches to source files and runs tests to measure kill rate; verdict is pass/warn/fail based on configurable thresholds; used by the mutation_test gate (opt-in, off by default)
Tool call/duration/repetition limit reached
Agent is stopped
`docs/`, `.swarm/outputs/`
string[]
Runs lint, secretscan, SAST, and quality budget in parallel (~15s vs ~60s sequential)
Read-only
string[]
Enforces complexity, duplication, and test ratio limits
Description
### Prefixed Agents Prefixed agents (e.g., `paid_coder`, `mega_reviewer`, `local_architect`) inherit defaults from their canonical base agent via `stripKnownSwarmPrefix`. The lookup order is: 1. Exact match for the prefixed name (if explicitly defined in user config) 2. Fall back to the canonical agent's defaults (e.g., `paid_coder` → `coder`) ```json { "authority": { "rules": { "coder": { "allowedPrefix": ["src/", "lib/"] }, "paid_coder": { "allowedPrefix": ["vendor/", "plugins/"] } } } } ``` In this example, `paid_coder` gets its own explicit rule, while other prefixed coders (e.g., `mega_coder`) fall back to `coder`. #### Selecting the primary agent in multi-swarm configs (`default_agent`) The top-level `default_agent` field controls which generated agents OpenCode treats as primary. **It is optional.** Behavior: - **Omitted** — every architect-role agent is primary. In a multi-swarm config that means each swarm exposes its own architect (`local_architect`, `mega_architect`, `paid_architect`, `modelrelay_architect`, …) as a selectable session default. This is the v7.0.0-compatible behavior and the recommended setup. - **Base role** (e.g. `"coder"`) — every generated agent whose canonical base role matches becomes primary (`local_coder`, `mega_coder`, …). - **Exact generated name** (e.g. `"local_architect"`) — only that agent is primary. - **Unknown / invalid value** — a one-time warning is logged and the resolver falls back to architect-role primaries (or the first generated agent if architects are disabled). The plugin never produces zero primaries when at least one agent exists. See [`docs/configuration.md`](docs/configuration.md) for the full table. ### Runtime Enforcement Architect direct writes are enforced at runtime via `toolBefore` hook. This tracks writes to source code paths outside `.swarm/` and protects `.swarm/plan.md` and `.swarm/plan.json` from direct modification. ### Configuration Override default rules in `.opencode/opencode-swarm.json`: ```json { "authority": { "enabled": true, "rules": { "coder": { "allowedPrefix": ["src/", "lib/", "scripts/"], "blockedPrefix": [".swarm/"], "blockedZones": ["generated"] }, "explorer": { "readOnly": false, "allowedPrefix": ["notes/", "scratch/"] } } } } ``` ### Rule Fields
string[]
`{ prose: 0.25, code: 0.4, ... }`
Post-coder typecheck for TS/JS, Go, Rust, C# (v6.29.2)
string[]
What It Does
`events.jsonl`
Production-grade test automation skills for 46 frameworks across E2E, unit, mobile, BDD, visual, and cloud testing in 15+ languages.
Audit a live web page in five phases (catalog, click, trace, classify, report) to identify mock data, hardcoded values, LLM-generated metrics, and broken endpoints. Outputs a markdown report with REAL/MOCK/LLM/HARDCODED/BROKEN/UNKNOWN verdicts per visible value.
X-ray any AI model's behavioral patterns — refusal boundaries, hallucination tendencies, reasoning style, formatting defaults. No API key needed.
Proves work with tests, diffs, and logs before sign-off.
Mobile app automation and verification for AI coding agents. CLI, MCP server, and typed Node.js API for iOS, Android, HarmonyOS, TV, web, macOS, and Linux.
AgentEval is the comprehensive .NET toolkit for AI agent evaluation—tool usage validation, RAG quality metrics, stochastic evaluation, and model comparison—built first for Microsoft Agent Framework (MAF) and Microsoft.Extensions.AI. What RAGAS, PromptFoo and DeepEval do for Python, AgentEval does for .NET
Cognitive Browser: The browser automation that thinks. Constitutional safety • Persona UX testing • Natural language interface • Self-healing selectors • Built for AI agents
Learn how to use the absolute-deflake Claude skill. Complete guide with installation instructions and examples.
Learn how to use the prompt-reliability-engine Claude skill. Complete guide with installation instructions and examples.
Learn how to use the expect Claude skill. Complete guide with installation instructions and examples.