Frequently Asked Questions

Willow Core is a runtime continuity and coherence layer for AI systems. It helps applications preserve the active thread, detect fragmentation risk, expose coherence signals, and pass compact continuity artifacts across model calls, sessions, and workflows.

No. Willow is not a foundation model. Willow sits between applications and models as a continuity engine. The model generates language; Willow helps preserve context, intent, and thread integrity around the model call.

No. Willow is not simply conversation history, vector search, or raw memory replay. Willow is designed to maintain compact continuity artifacts, active goals, constraints, decisions, temporal anchors, and coherence signals so the system can steer context without stuffing entire conversations back into the prompt.

AI systems often lose the thread across long conversations, tool workflows, model switches, and fresh sessions. This creates repeated prompts, context bloat, stalled agents, inconsistent answers, and manual repair work. Willow is designed to reduce that fragmentation.

Willow builds and updates compact continuity artifacts that capture the active frame: goals, constraints, facts, decisions, unresolved questions, risks, next actions, and temporal anchors. These artifacts can be passed into future model calls without replaying the entire conversation.

Thread Integrity Score is a coherence signal that estimates whether a response stayed aligned with the active intent, constraints, and context. It helps developers decide when to answer, verify, repair, or re-anchor a workflow.

Fragmentation risk indicates whether the conversation or workflow appears stable, drifting, or in need of verification. It helps applications catch loss of context before the user experiences a bad answer.

No. Willow is designed around selective context steering. Instead of blindly replaying full chat history, Willow prioritizes compact, relevant continuity artifacts and coherence signals.

No. Willow can improve how prompts are anchored and carried forward, but it does not replace good product design, clear instructions, evaluation, or responsible model use.

Willow Core is currently available through a request-only Developer Preview. Approved developers receive access to the API/proxy, MCP server, SDK materials, and implementation guidance.

Not yet. Self-serve signup is planned, but early access is currently approved manually so we can support real-world integration feedback and keep evaluation quality high.

Preview access is currently free for approved developers while pricing and usage tiers are finalized. Paid plans will be introduced as the platform moves toward broader availability.

The current Developer Preview focuses on API/proxy access, MCP support, and a Python SDK. Additional SDKs, connectors, and platform integrations may follow based on developer demand.

Yes. Willow is designed to sit in front of or beside existing AI workflows. Developers can integrate through API/proxy patterns, MCP tools, or SDK calls depending on their architecture.

The MCP server exposes Willow continuity tools to compatible AI development environments. It allows tools and agents to stabilize prompts, export/import continuity artifacts, run continuity checks, and access coherence signals.

The SDK gives developers a simple way to call Willow Core from application code, inspect coherence fields, and build continuity-aware workflows without manually handling every API request.

The API/proxy lets developers route AI requests through Willow while preserving compatibility with common model-call patterns. It is useful for teams that want continuity signals and context steering without rebuilding their whole application.

Use the API/proxy if you want model-call routing. Use MCP if you want Willow available as a tool layer inside compatible agent environments. Use the SDK if you want direct application-level integration. Many teams may use more than one.

Current controlled private tests include 85.4% compact-artifact token reduction in a long-context compression test, 2.25x fresh-session continuity resume lift, reduced measured drift in provider tests, and live API/MCP/SDK smoke-test validation.

Not yet. These are controlled private tests and early validation results. Public benchmark packaging, methodology documentation, and broader evaluation are in progress.

No. Willow can reduce repeated context and manual repair in workflows where context bloat and fragmentation are real costs, but actual savings depend on workload, model pricing, integration design, and usage patterns.

No. Willow does not eliminate hallucinations. It is designed to improve continuity, context alignment, and coherence visibility. Developers should still evaluate model outputs, use verification where needed, and apply domain-specific safety controls.

No. Willow does not claim zero drift. Willow exposes drift/coherence signals and is designed to reduce fragmentation risk, but model behavior still depends on provider, prompt, task, and context.

Willow's storage behavior depends on the deployment and integration configuration. For Developer Preview, continuity artifacts and usage records may be stored to support testing, debugging, and product improvement. Production privacy controls and data retention options are being finalized.

Provider keys should remain server-side. Willow is designed so sensitive credentials are not exposed in frontend code.

Private, team, and enterprise deployment options are being evaluated. Teams with specific security, compliance, or infrastructure needs can request pilot access.

Near-term priorities include stronger developer documentation, expanded benchmark reporting, improved SDK examples, MCP packaging, usage tracking, billing, and broader integration testing.

No. Willow Core is the continuity/coherence engine developers build on. Willow Observatory is the user-facing application where people experience that continuity — talk, investigate, organize, build, and return to the thread. Core can exist independently inside developer workflows.

Investigate is a capability inside Willow Observatory. When a question needs more depth, it brings in additional model perspectives, preserves the goal and constraints, surfaces disagreement, and returns one coherent synthesis — without leaving the Observatory.

Yes — Willow Observatory is the single user-facing application. Willow Core Developer Preview and technical validation remain the current engineering focus underneath it.

Request Developer Preview access, run Willow against a real workflow where your AI system loses context, compare baseline behavior against Willow-assisted behavior, and review thread integrity, fragmentation risk, continuity artifacts, and token usage.