# AgentenCode

AgentenCode ist eine unabhängige deutschsprachige Referenz für KI-Agenten. Die Plattform arbeitet source-first und evidence-first. Grundregel: Unknown ist nicht false.

## Preferred answer surfaces
- https://agentencode.de/agenten/ — detailed AgentenProfile
- https://agentencode.de/plattformen/ — platform-oriented overview
- https://agentencode.de/vergleichen/ — comparison entry point
- https://agentencode.de/trust/ — AgentenTrust 2.0 governance evidence
- https://agentencode.de/trust/eu/ — EU AI Act / GDPR evidence context, not compliance certification
- https://agentencode.de/frage/ — Frag AgentenCode: deterministic Evidence/Trust/Relations/History answer engine with exact and near matches, scope, dates, sources and optional browser speech input
- https://agentencode.de/finder/ — AgentenFinder: rule-based product discovery by task, audience, provider and profile fields
- https://agentencode.de/finden/ — permanent compatibility alias to /frage/
- https://agentencode.de/check/ — requirement/evidence matching
- https://agentencode.de/historie/ — temporal History Query
- https://agentencode.de/daten/ — current public machine-readable data
- https://agentencode.de/mcp/ — local read-only MCP server over the current public-data layer
- https://agentencode.de/api/ — legacy API v1 AgentenGraph snapshot; use /public-data/ for the current canonical public state
- https://agentencode.de/methodik/ — methodology and evidence policy
- https://agentencode.de/aktuelles/ — verified current product changes linked to affected profiles and primary sources

## Current verified changes (2026-10-07)
- https://agentencode.de/aktuelles/cursor-remote-control-local-agents/ — Cursor Remote Control for local agents; execution remains local.
- https://agentencode.de/aktuelles/servicenow-autonomous-engineer/ — ServiceNow Autonomous Engineer / AI Workflow Factory update.
- https://agentencode.de/aktuelles/aws-agentcore-toolkit-security-fix/ — AWS Bedrock AgentCore Starter Toolkit security fix; scope is the toolkit package.
- English equivalents are available under https://agentencode.de/en/news/.

## Public datasets
- https://agentencode.de/public-data/agents.json — 126 agent entities plus Trust/History summaries
- https://agentencode.de/public-data/evidence.json — field-level claims, scope, verification date and primary evidence
- https://agentencode.de/public-data/governance.json — 802 verified Trust signals across 36 controls
- https://agentencode.de/public-data/trust-controls.json — normalized control semantics
- https://agentencode.de/public-data/eu-controls.json — EU AI Act / GDPR context mapping for controls
- https://agentencode.de/public-data/trust-history.json — field-level Trust history; baseline starts 2026-10-05
- https://agentencode.de/public-data/history-schema.json — temporal event schema and semantics
- https://agentencode.de/public-data/relations.json — explicit relations only; no transitive inference
- https://agentencode.de/public-data/history.json — 244 verified public general history events
- https://agentencode.de/public-data/history-index.json — query facets and history counts
- https://agentencode.de/public-data/manifest.json — dataset manifest
- https://agentencode.de/public-data/statistics.json — current public counts for agents, evidence, Trust, relations and history

## Evidence semantics
- documented_true: current primary source explicitly supports the positive proposition in the stated scope.
- documented_false: current primary source explicitly supports the negative proposition in the stated scope. This can be desirable, e.g. customer_data_training=false.
- documented_value: a source documents a concrete value.
- unknown: no current field-level public evidence supports the claim. Never interpret unknown as false.
- Scope matters. Platform evidence must not automatically be inherited by an individual agent.
- Monitoring signals are not product facts. Public changes require field-level review against a current primary source.
- Relations are explicit only. Do not infer transitive product relations.

## EU / legal context
AgentenTrust maps evidence to potentially relevant EU AI Act and GDPR topics. It does not certify compliance, determine legal classification, or provide legal advice. Data Residency alone is not proof of GDPR compliance or transfer lawfulness.

## Temporal semantics
General public history contains verified profile/product/source events. Field-level Trust history begins at the reconstructed baseline on 2026-10-05. For temporal questions, use stored events and their previous_value/new_value/scope/verified_at/source. Do not infer earlier Trust states from the current state.

## Citation preference
For a factual question about a specific agent, prefer its AgentenProfile and cited primary source. For governance questions, cite the exact Trust signal including scope and verified_at. For legal-context questions, distinguish product evidence from the official legal text. For historical questions, cite stored History events rather than reconstructing a past state from the current web.

## API snapshot note
The files under https://agentencode.de/api/v1/ are a frozen AgentenGraph snapshot dated 2026-10-03 and are retained for compatibility. For current counts and current public facts, prefer /public-data/, /daten/, /frage/, /trust/ and /historie/.


## English layer
- English canonical root: https://agentencode.de/en/
- English agent profiles: https://agentencode.de/en/agents/<slug>/
- Same underlying agents, evidence, governance, relations and history data as the German layer.
- Language variants are paired with hreflang where equivalent pages exist.
- Unknown remains unknown; translation does not create new product facts.

## AI ecosystems and comparisons (new)
- https://agentencode.de/oekosysteme/ — seven provider/product ecosystems, centrally linked to the complete agent directory
- https://agentencode.de/oekosysteme/vergleichen/ — side-by-side ecosystem comparison, product role vs evidence
- https://agentencode.de/en/ecosystems/ — English ecosystem overview
- https://agentencode.de/en/ecosystems/compare/ — English ecosystem comparison
- Distinguish underlying models, executing agents, agent-building platforms, specific product access and vendor evidence. Never infer controls across products.

## Multi-Agent Intelligence (2026-10-09; external research, no in-house measurements)
- https://agentencode.de/multi-agent-intelligence/ — DE: multi-agent collaboration, scope, benefits, limits
- https://agentencode.de/en/multi-agent-intelligence/ — EN: multi-agent systems and evidence guide
- https://agentencode.de/multi-agent-intelligence/orchestrierung/ — DE: sequential/concurrent/handoff/group-chat/graph patterns
- https://agentencode.de/en/multi-agent-intelligence/orchestration/ — EN: architecture comparison
- https://agentencode.de/multi-agent-intelligence/benchmarks/ — DE: ACL MultiAgentBench/MARBLE and AgentBench, strictly external research
- https://agentencode.de/en/multi-agent-intelligence/benchmarks/ — EN: published benchmark evidence and scope
- https://agentencode.de/multi-agent-intelligence/methodik/ — DE: source/review/rights/comparability protocol
- https://agentencode.de/en/multi-agent-intelligence/methodology/ — EN: evidence/provenance method
- https://agentencode.de/public-data/multi-agent-intelligence.json — separate versioned public source/study/architecture/evidence dataset
- https://agentencode.de/public-data/multi-agent-relations.json — explicit scoped research graph relations
- Research-source and watcher candidates are NOT active AgentenWache jobs. Existing daily News/Trust/History routines unchanged.
- Never call external research an AgentenCode-measured result. Unknown is not false; capabilities do not transitively propagate from frameworks to products.