Ecosystem overview

Use the homepage as the system front door for memory, signals, models, tasks, and risk.

Keep the original graph visible while adding signal, task, evidence, and AI context around it.

Route deep work into Agents, Memory, Signals, and Operations without overloading one page.

Let OneAI read scoped context and produce reviewable proposals, not direct commits.

Pass every signal through quality, evidence, and memory-candidate gates.

Turn memory and signals into task drafts, owner checkpoints, and coordination moves.

Keep user memory scoped, consented, revocable, auditable, and free of raw secrets.

High-quality signal with enough strength, confidence, and freshness for review.

No candidate lifecycle items need review.

Review as a coordination recommendation before creating or updating a mission.

Private memory, secrets, raw vault exports, and sensitive identifiers must stay out of model context.

Mission backlog pressure

agent signal routed to background-context.

WriteIntent creation exists but remains gated by database mode, backup, and ONEFIELD_P6_WRITES_ENABLED. Use /api/production/write-intents/dry-run before any real POST, then create only targeted intents.

Session signing is production-ready, but durable claim and role request tables still need an additive migration. Add additive UserClaim, RoleUpgradeRequest, and Invite tables before enabling register/claim writes.

OneAI live proposal infrastructure is ready or reachable. Use /api/oneai/health?probe=true, then POST /api/oneai/live-proposals from an admin/steward session.

4 signal sources are configured/live; 0 sources report error. Keep exchanges public-readonly; add scheduler/webhook replay checks before treating feeds as realtime production.

FieldMemoryVersion exists, but full memory create/revoke/export lifecycle still needs dedicated routes and policy checks. Enable memory writes target-by-target after backup; start with versioned memory create and export APIs.

Basic in-memory rate limiting and audit logging exist; tenant, billing, usage ledger, and external observability are not complete. Add additive UsageEvent, Tenant, ApiKey, BillingAccount, and SecurityReview tables before opening external customers.

Security launch review requires admin/steward context. Run the launch checklist and rotate any exposed keys before opening real external access.

Review the blocked boundary and keep the capability or candidate blocked until a human approves a new path.

Candidate, confirmation, decay, conflict, and archive states make memory governable before persistence.

Agent Context Runtime defines who can see which memory, signal, risk, evidence, and approval surfaces.

6 fused signal themes explain how market, social, agent, and community inputs become coordination context.

Approval gradient expands AI autonomy for organization and recommendation while keeping durable state changes human-governed.

Coordination Graph v2 treats people, agents, models, signals, missions, memory, risks, and approvals as one explainable coordination surface.

External AI Memory API exposes OneField as infrastructure without handing agents private memory or mutation authority.

Confirm a fresh backup and restore drill outside the app before enabling P8.1 writes.

Operator console can approve, reject, request evidence, inspect rollback, and confirm backup in a future write path.

Every consent must be revocable. Revoked or expired consent must deny future scoped context requests.

Add additive memory lifecycle persistence for versions, confirmations, revocations, conflicts, decay, and archive state.

1 write-intent previews can shape durable memory candidates.

Create per-agent identities for OneAI, Telegram, OneMission, external AI, and operator tools with scoped access grants.

6 gateway capabilities are allowed in read/draft scope.

Continuously fuse OKX, Binance, X, Telegram, OneMission, and agent logs into attributed signal themes.

8 signals across 4 configured/live sources.

Add an operator console to approve, reject, request evidence, attach backup confirmation, and inspect rollback hints.

1 AI submit-readiness items are visible before any POST.

Add API keys, SDK docs, rate limits, usage logs, scoped memory grants, and developer onboarding.

External AI endpoints are documented as scoped read-only protocol surfaces.

Let AI auto-organize and recommend in low-risk zones while routing durable changes to manual confirmation.

10 execution readiness items are classified into autonomy lanes.

Publish a protocol surface for cross-community, cross-agent, and cross-model coordination memory.

OneField exposes stable read-only contracts for memory, evidence, coordination, gateway, and operator packs.

Give agents long-lived context about trust, relationships, missions, signals, evidence, and human-approved decisions instead of raw chat history.

Normalize Binance, OKX, X, Telegram, OneMission, chain, and agent-log signals into evidence-backed memory candidates.

Turn field state into daily priorities, builder follow-up, relationship work, mission pressure, and reviewable action drafts.

Check provenance, permissions, policy, and human confirmation before any future mission, graph, memory, or asset-affecting action.

Let agents read, reason, cite evidence, and draft recommendations while humans retain write authority and final judgment.