oss

Return one Base120 model by code.

List Base120 models, optionally filtered by transformation code.

Search Base120 model codes, names, and definitions.

Format a deterministic local prompt for a Base120 model and question.

Initialize a new BIF ingestion session. Returns session_id, Phase 1 prompt template, and pre-ingestion checklist.

Get the prompt template for a specific BIF phase and batch. Returns structured prompt with execution steps, quality checks, and file naming conventions.

Validate a completed batch against BIF quality criteria. Checks metadata header, structure, code examples, tables, token budget, and more.

Check progress of a BIF session. Without session_id, lists all sessions. With session_id, shows detailed progress including phase coverage and next batch.

List available BIF domain templates (AI/ML Platform, Cloud Platform, Programming Framework, SaaS Evaluation, and the generic template).

Get a one-call overview of all governance primitives: kill switch mode, circuit breaker state, cost budget status.

Get detailed kill switch state: current mode, engagement history, critical tasks list.

Engage the kill switch. Modes: HALT_NONCRITICAL, HALT_ALL, EMERGENCY. Requires confirm=true.

Disengage the kill switch and return to normal operations.

Get circuit breaker state: CLOSED (healthy), OPEN (tripped), or HALF_OPEN (testing recovery).

Check current daily spend against budget caps. Returns ALLOW, WARN, or DENY decision with rationale.

Record an API usage event for cost tracking.

Query the governance audit log by intent_id or task_id.

Generate compliance evidence mapped to SOC2, GDPR, and OWASP controls from governance audit trail.

Run governance health probes and return overall status.

Run CI contract locally before push. Scopes checks to the diff.

Check branch readiness before PR creation. Verifies branch freshness, conflict risk, scope, and PR metadata.

Claim a review aspect on a PR to prevent duplicate work. Aspects: security, logic, tests, docs, style, ci, performance, breaking.

Query the review coverage matrix for a PR. Shows which aspects have been reviewed.

Watch CI status for a PR. Polls at the given interval and returns

Check if origin/main advanced relative to local. Reports stale branches.

Detect stale branches and optionally rebase. Without --execute, reports which branches can be cleanly rebased. With --execute, performs the rebase.

Read recent coordination bus messages. Optionally filter by agent or message type.

Post a message to the coordination bus. Uses bus_writer for flock-safe append.

Search bus messages by text content across all fields.

Get bus statistics: total messages, date range, top agents, message type breakdown.

List all agents that have posted to the bus with message counts, first/last seen, and top message types.

Full-text BM25 search across all cognitive ledger entries. Returns ranked results with relevance scores.

Query ledger entries by structured filters: agent, type, scope, tags, or time range.

Append a new entry to the cognitive ledger. Returns the entry ID.

Get ledger statistics: entry count, agent breakdown, type breakdown, index health.

Get session boot context — high-value recent entries for agent initialization. Returns the most relevant entries for starting a new work session.

Rebuild the BM25 full-text index from the ledger. Run after bulk imports or if search quality degrades.

Record an agent action for drift analysis

Snapshot current action distribution as the baseline for drift comparison

Run Jensen-Shannon drift + entropy gaming detection for an agent

Record an agent action and classify it against convergent goal patterns

Check an agent for instrumental convergence (Bostrom 2014)

Get per-convergent-goal scores for an agent

Run full NIST AI RMF governance check (kill switch + identity + circuit breaker + cost)

Get full governance lifecycle status snapshot (all NIST functions)

Register a new AI variant in the lineage graph

Get variant details, children, and optionally full ancestry

Detect parent-to-child fitness drift across all variants

Request human approval for an agent action. Creates a PENDING request

Check the status of an approval request. Auto-expires stale requests.

Approve a pending approval request. Requires reviewer identity.

Deny a pending approval request. Requires reviewer identity.

Cancel a pending approval request (typically called by the requesting agent).

List all pending approval requests awaiting human decision.

List all approval requests (newest first).

Get aggregate approval statistics: counts by status and risk level.

One-call gate: if risk_level is below threshold, auto-approves. Otherwise creates a request and blocks until a decision is reached or timeout expires. Returns {allowed, status, request_id, reason}.

Manually trigger expiration of all stale (timed-out) pending requests.

Map governance evidence to NIST CSF 2.0 and AI RMF categories. Reads governance JSONL audit logs and maps each tuple type to the relevant NIST subcategories, returning coverage percentages and gaps.

SOC 2 Type II readiness assessment. Scores governance evidence against all five trust service criteria: Security, Availability, Processing Integrity, Confidentiality, and Privacy. Returns per-criteria readiness.

Cross-reference controls across ISO 27001, ISO 42001, NIST CSF 2.0, NIST AI RMF, and SOC 2. Returns a mapping table showing which controls have governance evidence and coverage percentages per framework.

Run STRIDE threat analysis on a set of agent interactions. Evaluates each interaction for Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, and Elevation of Privilege threats.

Export a complete compliance evidence package as structured JSON for auditors. Combines evidence from all governance primitives mapped to

Register an agent with a trust tier

Look up agent status and trust tier

List all registered agents

Check if an agent is a valid sender

Create a signed delegation token binding issuer to subject for ops

Validate a delegation token by token_id

Check least-privilege: is requested_op in token's ops_allowed?

Advance the Lamport clock for a local event

Advance the clock on receipt of a remote timestamp

Compare two Lamport timestamps for causal ordering

Check if proposed velocity/force/jerk values are within kinematic limits

Retrieve configured kinematic limits and caution/emergency thresholds

Get effective velocity limit for a given safety mode

Evaluate pHRI safety mode from distance/collision sensor data

Get configured pHRI proximity thresholds and mode boundaries

Check multiple sensor readings at once, returns per-reading safety mode

List all available Base120 mental models

Get details of a specific Base120 mental model by code

Generate a specialized system prompt for a Base120 mental model

Validate a JSON value against a JSON Schema (Draft 2020-12 subset)

Validate a dict against a schema, returns (valid, errors)

Announce a task for bidding via Contract Net Protocol

Submit a bid on an announced task

Evaluate bids and select a winner

Get phase, bid count, and winner for an announcement

Count all contract announcements by phase

Create an isolated sandbox for an agent with policy constraints

Check if an action is allowed within a sandbox.

Validate agent output against sandbox rules (no secrets, no blocked content).

Get full sandbox state: kill switch, circuit breaker, cost, actions.

Tear down a sandbox and emit an audit receipt.

Read the current shared cognitive state. Returns version, active agents, claimed files, decisions, sprint, and flags as JSON.

Update a specific agent's status in the shared state. Creates the agent entry if it doesn't exist.

Claim exclusive access to a file. Prevents other agents from modifying it concurrently. Fails if already claimed by a different agent.

Release exclusive file access. Only the claiming agent can release (or use agent_id='*' to force).

Read or set flags in the shared state. If value is omitted, returns the current value of the flag. If value is provided, sets the flag.