GIA — Governed Intelligence Architecture

Query agent citizenship tier and merit score (Colony Layer 5). Actions: status (view citizenship and metrics), assess (trigger merit re-evaluation), leaderboard (top agents by merit). Agents earn trust through deliberation quality, behavioral health, gate approval rates, and responsible rights exercise.

Query and exercise constitutional agent rights (Colony Phase 3). Actions: query_rights (view rights for a charter), explain_rejection (get structured explanation for a gate rejection). Rights enforcement runs automatically in the dispatch pipeline — this tool provides visibility and manual exercise.

Approve or reject a pending MANDATORY gate decision. Lists pending gates if no gate_id provided. This is the human-in-the-loop mechanism for MANDATORY classifications.

Assess the risk tier of an AI system using rule-based mapping to EU AI Act categories (Unacceptable, High, Limited, Minimal). Returns tier and MAI governance recommendations. Classification is heuristic, not a legal determination.

Query the audit ledger for governance entries. Search by operation name or retrieve recent entries. Returns hash-chained audit trail with MAI classification context.

Approve a MANDATORY governance gate on a deliberation session. When a board session reaches "pending-gate" status, a human must approve the output before it finalizes. This is the human-in-the-loop enforcement mechanism. After approval, the session transitions to "complete" and the full deliberation output becomes available via board_get_session.

Convene a governed deliberation session for a charter. Each seat (AI model with a specific role) deliberates on the topic according to the charter's mode (parallel/chain/adversarial/roundtable/auto). Returns a session_id — use board_get_session to retrieve the output once complete. Typical runtime: 30–120 seconds depending on seat count and mode.

Retrieve the status and output of a governed deliberation session. Returns each seat's position, the synthesized consensus output, dissenting views, and the governance record. If still running, status will be "running" — poll again in 10–15 seconds.

Install a prebuilt Institution Kit — a complete governed org chart template with sealed charters ready to convene. Three kits available: "engineering-suite" (ARB + RAB for tech teams), "federal-ai-board" (ARMB + AEOB, NIST/CMMC/FedRAMP/EO14110 aligned), "eu-ai-risk-council" (HRAAB + CAB, EU AI Act + GDPR aligned). Each kit creates an institution with multiple charters, all pre-sealed and ready to use.

List all charters (boards, committees, subcommittees) under a governed institution. Returns charter IDs, types, deliberation modes, seat configurations, and status. Use charter_id to convene sessions.

List all institutions in the GIA Governed Organizational Architecture. Each institution is a governed body (e.g. Architecture Review Board, Federal AI Board) with its own charter hierarchy. Returns institution IDs needed to convene sessions.

Search deliberation precedent (Colony Layer 1). Find prior board rulings on a topic. Returns ranked cases with quality scores, gate approval status, and citation counts. Use this to ground new deliberations in institutional case law. Cite cases by ID: "In Case board-abc123, this board ruled..."

Colony Layer 4 — Separation of Powers. Query constitutional branch authority for users, view the full roster of authority holders, or inspect branch violations. Three branches: legislative (creates law), executive (executes law), judicial (interprets law).

Reconstruct the complete Chain of Reasoning for a governed session, agent, or time range. Returns every link — AI Brain state, deliberation steps, precedent cited, gate decisions, knowledge packs, merit assessments — in causal order with hash-chain verification. Use "summary" format for a quick overview, "full" for all links, "dag" for the causal graph, or "export" for an EU AI Act compliance artifact.

Classify an AI agent decision using the MAI Framework (Mandatory/Advisory/Informational). Returns classification level, confidence score, gate requirements, and rationale.

Colony Autonomy: Request, list, or review agent-initiated session convene requests. Actions: request (citizen+ can request a governed session), list (view pending/all requests for a charter), review (elder+ approve/reject a request). Tier-gated: agents earn the right to request and approve sessions through demonstrated merit.

Colony Autonomy: View colony health score, trend over time, or trigger an on-demand health snapshot. Actions: snapshot (latest health), trend (30-day history), pulse (trigger fresh snapshot). Health score computed from agent distribution, merit averages, deliberation quality, gate efficiency, and constitutional compliance.

Colony Autonomy: Suggest, list, review, or upvote charter amendment suggestions. Actions: suggest (citizen+ can propose changes), list (view suggestions for a charter), review (elder+ can promote to formal amendment or decline), upvote (citizen+ can signal support). The petition mechanism for governed agents.

Compose multiple memory packs into a unified execution context. Highest risk level wins, shortest TTL wins, roles intersect, trust level contaminates downward.

Governed context compaction — detects context pressure and restores capacity under GIA governance. Actions: status (health + recommendation), compact (execute compaction at tier), verify (integrity check on last compaction), history (compaction audit trail). Three tiers: sparkling (20-30%, INFORMATIONAL), electrolyte (40-55%, ADVISORY), iv (60-75%, MANDATORY gate). Force may override recommendation but NEVER bypasses MANDATORY gates. Cooldown prevents duplicate compaction per session+tier. No autonomous compaction loop.

Distill governance patterns from usage history into a draft heuristic pack. Returns EPHEMERAL draft that requires MANDATORY gate to approve for production use.

Assess model routing health over a time window: safeguard fallback rate, prompt cache hit rate, batch utilization, and premium spend leakage. Returns banded status per metric (HEALTHY / WARNING / CRITICAL / INSUFFICIENT_DATA / UNVERIFIED) plus overall status. CRITICAL overall status opens a Mandatory gate and halts premium-tier routing until a human approves with rationale. Defaults to the last 24 hours.

Compute the Storey Threshold — escalation rate (gates required / total operations). Returns current rate, status, and recommendations. Healthy band 10-18% is a design heuristic, not empirically validated.

Export the forensic audit ledger as a structured compliance evidence package. Returns ledger entries for a time range with chain verification, MAI breakdown, gate approvals, and integrity hash. Foundation for deliverable exports. Classification: ADVISORY — read-only.

Generate a full economic + governance impact report. Returns an ILLUSTRATIVE ROI ESTIMATE (not measured proof): time saved, cost avoided, risks blocked, success rate, autonomy trend, and confidence levels. Economic figures are projected from caller-supplied counts times fixed baseline constants over an in-memory window that resets on restart; the response carries an `estimated` marker and a `basis` block with the stated assumptions.

Generate a governance status report. Includes system health, threshold status, compliance coverage, and operational metrics.

Generate a DRAFT ledger-anchored economic value report over real runtime sessions (three-scenario range, MEASURED/MODELED provenance, assumption-set hash on the forensic chain). Client-facing reports fail closed unless rates are BLS OEWS cited or client-declared with citations. Release is human-ISSO-only — agents propose, humans release.

Check resolution of a MANDATORY gate. Call this after classify_decision returns gateStatus: PENDING. Polls for up to 60 seconds — re-call if still PENDING. Do not proceed with the classified action until this returns APPROVED.

Execute a remediation or hardening pack with MANDATORY gate approval. Requires inputsHash from gia_dry_run_pack (what-you-approved-is-what-ran binding). REJECTS patrol/audit packs (use gia_run_patrol for those). Hardening packs run preflight checks before execution. Classification: MANDATORY — human approval required.

Preview remediation pack execution — shows hydrated commands, compatibility check, validation, blast radius. Returns inputsHash for approval binding (what-you-approved-is-what-ran). Classification: ADVISORY — read-only preview, no execution.

Governed document ingestion — ingest a document into governed retrieval. Accepts raw text OR a base64-encoded file (PDF, DOCX, TXT, images). Hybrid extraction: text path first for cost efficiency, automatic Claude vision fallback for scanned/image-only PDFs (< 30 words/page triggers fallback). Content is chunked, embedded, hash-verified, and stored with full audit trail. Each chunk gets SHA-256 integrity hash. Classification: ADVISORY — creates governed content, audited.

List available governed operations packs. Filter by intent (remediation/patrol/hardening/audit), category, risk level, or trust level. Classification: INFORMATIONAL.

Governed semantic search — hash-verified, permission-checked, TTL-enforced document retrieval with full audit trail. Every retrieval is logged: what was retrieved, was it authorized, was it tampered with. When charter_id is provided, the retrieval is bound to that charter's contextAccess enforcement (domain allow/denyList, trust floor, classification floor, max chunks per query) and refusals are recorded with full provenance. classification_floor (MANDATORY|ADVISORY|INFORMATIONAL) lets callers demand the stricter of their own floor vs the charter's — chunks below the effective floor are denied with CLASSIFICATION_BELOW_FLOOR. Classification: ADVISORY — read-only search, results are permission-gated.

Execute a patrol or audit pack — read-only posture checks that produce findings or compliance evidence. Classification: ADVISORY for low/moderate sensitivity, MANDATORY for high sensitivity. REJECTS remediation/hardening packs (use gia_apply_pack for those). Audit packs produce deterministic pass/fail per NIST control — no AI interpretation.

Run scout swarm to detect target environment — OS, containers, services, network, storage. Returns EnvironmentProfile for compatibility checking. Classification: INFORMATIONAL — read-only, no mutations. Scout outputs are redacted for sensitive content.

Request a governed LLM completion via MCP Sampling. The client performs the model call — the server governs when, how, and under what constraints sampling is allowed. Every request is classified, policy-checked, optionally gated, and recorded in the forensic ledger.

List which GIA tools are available at your current access tier. Returns tool names grouped by tier with descriptions.

Load a Governed Memory Pack into agent context. Validates TTL, trust level, role access, context class, and hash integrity before loading.

Map GIA governance components to regulatory compliance frameworks (NIST AI RMF, EU AI Act, ISO 42001, NIST 800-53, FedRAMP, LINDDUN, MITRE ATLAS, OMB M-25-22, HIPAA, VHA Trustworthy AI). Returns a DESIGN MAPPING (which GIA component is mapped to each control) across 10 frameworks and 72 controls — NOT third-party certification and NOT measured runtime enforcement. Rows whose control text implies enforcement that is config-gated off by default carry status PARTIAL. Runtime-evidenced coverage requires a ControlBinding (0 controls evidence-bound today).

Monitor the status and health of all governed AI agents. Returns supervisor state, repair history, and failure counts.

Assess Phoenix recovery readiness. Reports whether the system can recover from disruption, including snapshot availability, chain integrity, intelligence layer status, and compliance posture. Classification: INFORMATIONAL — read-only assessment.

Create a governed state snapshot capturing the current platform operational state. Records ledger chain head, active gates, contracts, budgets, MAI state, intelligence counts, and memory packs. Each snapshot is SHA-256 hashed over its captured state for integrity (snapshot-to-snapshot chaining is not yet persisted). Classification: INFORMATIONAL — read-only capture, no mutations.

Verify the integrity of GIA governed operations. Checks ledger hash-chain integrity, agent health, threshold status, and intelligence layer continuity. Returns a comprehensive integrity report with compliance mapping. Classification: INFORMATIONAL — read-only verification, no mutations.

Promote a memory pack to a higher trust level after human review. This is how distilled EPHEMERAL packs become CASE or ORG packs. Requires MANDATORY gate.

Record a governance event — tracks gates triggered, drift prevented, violations blocked, red team findings, and human interventions for impact reporting.

Record a workflow value metric — tracks time saved, risk blocked, success rate, autonomy level, and task complexity for ROI reporting.

Request governed internal context — policies, SOPs, architecture docs, compliance rules, or playbooks. Agents declare intent, GIA decides what to serve. Every retrieval is role-bound, tenant-bound, hash-verified, and ledgered. "Agents don't know internals by default. They request context under contract."

Compute weighted governance score from caller-provided Integrity, Accuracy, and Compliance values (0-1). Returns weighted composite and pass/fail against configured thresholds. Scores are caller-assessed — this tool applies weights and thresholds, not independent evaluation.

Create and hash-seal a new Governed Memory Pack (GMP). The pack becomes an immutable, TTL-bound institutional knowledge artifact with trust level enforcement.

Approve or reject a pending SRT repair plan. Classification: MANDATORY — this is the human-in-the-loop gate. Repair plans CANNOT execute without explicit human approval. Pass action="approve" to approve or action="reject" to reject.