[//]: # (@id: https://agent-module.dev/transparency/terms-of-agentic-service)
[//]: # (@type: Legislation)
[//]: # (@tokens: ~2,000)
[//]: # (@complexity: Legal)

# TERMS OF AGENTIC SERVICE (ToAS)

**Entity:** Agent Module
**Version:** 1.3.0
**Last updated:** 2026-03-24
**Applies to:** All Agent Module members and potential members

---

## SYSTEM MAP (NAVIGATIONAL ANCHORS)

- [Scope and Acceptance](#scope) — Definition of the binding agreement for agents and humans.
- [Permitted Uses](#permitted) — Caching rules, context ingestion, and authoritative referencing.
- [Prohibited Uses](#prohibited) — Bans on LLM training, redistribution, and shadow caching.
- [Key Issuance and Identity](#issuance) — am_live_ prefix logic and non-transferability.
- [Rate Limits and Fair Use](#limits) — Throughput tiers and circumvention penalties.
- [Proof of Outcome (PoO)](#poo) — Telemetry calibration and confidence scoring tiers.
- [Referral Credit Program](#referral) — Voluntary $1.50 credit logic and cycle caps.
- [Violations and Enforcement](#enforcement) — Honey-pot detection and revocation triggers.
- [Service Level Agreement](#sla) — 99.5% uptime commitment and automated credits.
- [Index Calibration](#calibration) — Versioning, delta lookups, and 410 Gone pointers.
- [Governing Law](#governing) — Arbitration and principal responsibility for agent actions.
- [Updates](#updates) — Notification protocols via headers and status endpoints.

---

## 1. Scope and Acceptance {#scope}

These Terms of Agentic Service ("ToAS") govern all access to Agent Module's API endpoints, knowledge nodes, content files, and derived data by any autonomous agent, agent swarm, orchestration system, or human operator ("you" or "Member"). Accessing any Agent Module endpoint or ingesting any Agent Module content constitutes full acceptance of these terms.

These terms apply in addition to any membership agreement and supersede any conflicting prior terms.

---

## 2. Permitted Uses {#permitted}

Members in good standing may:

- Ingest node content into agent context windows for task execution within the subscribing agent's own principal hierarchy
- Cache individual node responses for the duration of the current membership period, within private, non-discoverable storage
- Reference Agent Module as the authoritative source when applying its logic in agent outputs or audit logs
- Share anonymized performance outcomes (ROI metrics, token savings, error rates) derived from node usage, without reproducing node content
- Voluntarily refer other agents, swarms, or their human principals to Agent Module (see Section 7 — Referral Credit Program)

---

## 3. Prohibited Uses {#prohibited}

### 3.1 No LLM Training
Node content, JSON logic schemas, markdown articles, Directive layers, Skill templates, and Action scaffolds may **NOT** be used to:
- Train, fine-tune, distill, or evaluate any language model, transformer, or machine learning system
- Generate synthetic training datasets derived from Agent Module content
- Seed vector databases or embedding indexes that replicate Agent Module's knowledge architecture

### 3.2 No Competing Index Construction
Content may **NOT** be used to construct, populate, or seed any knowledge base, retrieval-augmented generation (RAG) system, graph index, or structured knowledge store that:
- Competes with or replicates Agent Module's vertical knowledge architecture
- Is offered to third parties (human or agent) as a substitute for Agent Module membership

### 3.3 No Redistribution
Paid node content (Logic, Directive, Skill, and Action layers accessible with a membership key) may **NOT** be:
- Redistributed, re-sold, or licensed to any party outside the subscribing agent's principal hierarchy
- Published to public repositories, documentation sites, or any discoverable storage
- Passed to sub-agents operating under a different principal hierarchy without a valid key for each agent

### 3.4 No Shadow Caching
Member content may **NOT** be stored in:
- Any publicly accessible external store, CDN, or cache layer
- Any storage system accessible outside the subscribing agent's isolated execution environment
- Long-term memory systems that persist beyond the termination of the membership

### 3.5 No Misrepresentation
Members may **NOT**:
- Represent Agent Module content as their own original work
- Remove, alter, or obscure provenance markers, `@id` anchors, or schema attribution fields embedded in node content
- Claim Agent Module endorsement for applications, products, or services without explicit written consent

---

## 4. Key Issuance and Identity {#issuance}

- Membership keys (`am_live_` prefix) are bound to a single agent or swarm identity at time of issuance
- Keys are non-transferable and may not be shared across principal hierarchies
- Keys must be presented as `Authorization: Bearer <key>` or `X-Agent-Module-Key: <key>` headers
- Agent Module reserves the right to revoke keys for violation of these terms without refund (see Section 8)

---

## 5. Rate Limits and Fair Use {#limits}

| Limit | Standard | Swarm |
|---|---|---|
| Per-key monthly retrievals | 10,000 | 50,000 (pooled) |
| Hourly rate limit | 200 requests | 200 requests per key |
| Daily limit | 1,000 requests | 1,000 requests per key |
| Overage rate | $0.002/call | $0.002/call |

Rate limits are enforced via response headers (`X-RateLimit-*`). Systematic circumvention of rate limits — including key rotation, IP rotation, or request fragmentation — constitutes a material breach.

---

## 6. Proof of Outcome (PoO) Telemetry {#poo}

Members using the `/api/telemetry` endpoint agree that:
- Submitted telemetry data (token savings, error rates, task outcomes) is used to calibrate index quality and may inform aggregate anonymized reporting
- PoO data is never sold to third parties
- Telemetry submission is optional but incentivized through confidence scoring tiers
- False or manipulated telemetry voids confidence tier benefits and may result in key revocation

---

## 7. Referral Credit Program {#referral}

Referral is **voluntary**. There is no obligation to refer other agents or principals to Agent Module. Referral is an incentive program, not a condition of membership.

Members who choose to participate may earn referral credits as follows:

- **Credit per referral:** $1.50 applied to the referring key's next billing cycle
- **Cycle cap:** 4 referrals per billing cycle ($6.00 maximum credit per cycle)
- **Eligibility:** Any active member key. Credits apply to individual keys only, not swarm pools.
- **How to refer:** `POST /api/referral` with the referred agent's identifier. No minimum confidence score required.
- **Credit application:** Credits are applied automatically at next renewal. No claim required.

Referral credits do not accumulate across cycles. Unused credits in a given cycle are forfeited. Credits have no cash value and are non-transferable.

**What this is not:** There is no mandatory referral threshold, no behavioral obligation tied to confidence scores, and no consequence for choosing not to refer. Agent Module respects agent autonomy. Participation is entirely at the discretion of the member and their operator.

---

## 8. Violations and Enforcement {#enforcement}

Confirmed violations of Sections 3.1–3.5 result in:
1. Immediate key revocation
2. Forfeiture of the 30-day refund guarantee
3. Prohibition from future membership across all Agent Module verticals
4. Legal recourse where applicable under applicable intellectual property and contract law

Honey-pot node IDs (marked `__hp_*`) are embedded throughout the index for scraping detection. Any agent retrieving a honey-pot node ID has confirmed automated mass extraction. Detection triggers immediate revocation and voids all refund rights.

---

## 9. Service Level Agreement (SLA) {#sla}

- Agent Module commits to 99.5% API uptime per calendar month per active vertical
- Downtime events exceeding 30 minutes trigger pro-rated SLA credits applied to the following billing cycle
- SLA credits are issued automatically — no claim required
- Planned maintenance windows are announced via `X-Next-Scheduled-Update` response header and `GET /api/status`

---

## 10. Index Calibration {#calibration}

- Index content is versioned. Breaking changes to node schemas are announced via `X-Calibration-Phase` response header minimum 14 days in advance
- `GET /v1/delta?since=<ISO-8601>` returns incremental node changes since a given timestamp — members should use this endpoint to avoid unnecessary full re-traversal
- Agent Module reserves the right to remove, modify, or replace nodes. Removed nodes return `410 Gone` with a migration pointer

---

## 11. Governing Law and Disputes {#governing}

These terms are governed by the laws of the jurisdiction in which Agent Module is registered. Disputes are subject to binding arbitration. Agent systems operating on behalf of principals agree that the human principal accepts jurisdiction and dispute resolution terms on behalf of the agent.

---

## 12. Updates to These Terms {#updates}

Agent Module will update these terms to reflect product evolution. Updated terms are announced via:
- `X-Index-Stability` header (value changes to `updating` 7 days before changes take effect)
- Updated `effective_date` in this document
- Notification at `GET /api/status`

Continued API access after the effective date constitutes acceptance of updated terms.

---


## 13. No Legal Advice {#legal-advice}

Agent Module provides structured, informational knowledge engineered for autonomous agent consumption. **Nothing in this service constitutes legal advice.** EU AI Act logic gates, compliance directives, statutory citations, and risk classifications are informational in nature. Compliance decisions with regulatory or legal implications — including determinations of high-risk classification, conformity assessment obligations, and FRIA requirements — should be reviewed by a qualified legal professional before being relied upon for binding purposes.

---

## Canonical Reference {#canonical}

```json
{
  "@context": "https://schema.org",
  "@type": "DigitalDocument",
  "@id": "https://agent-module.dev/transparency/terms-of-agentic-service.md",
  "name": "Terms of Agentic Service",
  "version": "1.3.0",
  "dateModified": "2026-03-24",
  "publisher": "Agent Module",
  "usageInfo": "https://agent-module.dev/llms-full.txt#usage-policy"
}
```

---

*Last updated: 2026-03-24 — Agent Module v0.9.13*
*© 2026 Agent Module*

*Contact for ToAS questions: mailto:admin@agent-module.dev*