Question 1

Which local models should I run, and how does hardware affect SecureContext?

Accepted Answer

SecureContext degrades gracefully by hardware tier. Minimum: any CPU with ~2 GB RAM runs only nomic‑embed‑text embeddings, providing hybrid BM25 + vector search and core memory features. Mid tier (8–16 GB GPU or Apple Silicon 16 GB) adds qwen2.5‑coder 14b or phi4 14b for event extraction, contradiction adjudication, and semantic file summaries. Full tier (24 GB+ GPU) runs phi4 14b and gpt‑oss 20b together, enabling all features at interactive latency. The required embedding model (nomic‑embed‑text) runs everywhere.

Question 2

What do I need to run it?

Accepted Answer

Node 20+ and (recommended) Docker for the bundled PostgreSQL + Ollama stack. The one‑command installer sets everything up in about five minutes. A SQLite fallback runs with zero infrastructure.

Question 3

Can multiple Claude Code sessions work on the same project without conflicts?

Accepted Answer

Yes. Parallel sessions claim tasks atomically from a work‑stealing queue, coordinate via typed broadcasts (ASSIGN/STATUS/MERGE/REJECT), and maintain private per‑agent memory namespaces plus a shared pool. Hierarchical team structures are supported.

Question 4

Are Claude Code skills safe to install?

Accepted Answer

Filesystem skills run with your permissions and are not scanned by Claude Code. SecureContext adds safety by scanning skill scripts with an AST gate, verifying them with HMAC before execution, automatically quarantining failures or post‑admission changes, and logging every admission decision in a verifiable chain.

Question 5

Does SecureContext send my code or data to the cloud?

Accepted Answer

No. All memory, embeddings (via Ollama nomic‑embed‑text), search, summarization, and the audit chain run locally. SecureContext works fully offline and incurs $0 cost when idle.

Question 6

Can it manage a long, multi‑week project?

Accepted Answer

Yes. The delivery layer lets you define a program with phases (zc_program). Open phases carry an acceptance checklist; closing a phase generates a checkpoint document. If the orchestrating agent dies mid‑phase, a fresh agent can reconstruct the full delivery state from two calls and resume work.

Question 7

How much does SecureContext reduce Claude Code token costs?

Accepted Answer

In a 10‑session project SecureContext reduces context‑overhead by about 87 %, shrinking session restore from 20–50 k tokens to roughly 1.5–4 k tokens. Long‑lived projects keep each recall to ~4 k tokens instead of ~47 k, and file queries use 15–50× compressed semantic summaries.

Question 8

How do I give Claude Code persistent memory across sessions?

Accepted Answer

Install SecureContext using the one‑command Quick Start, then have agents call `zcremember` to persist decisions and gotchas and `zcrecall_context({focus: "current task"})` at session start to restore them. Memory is stored in PostgreSQL, surviving restarts, compactions, and reboots.

Question 9

How is it different from competitors?

Accepted Answer

SecureContext differs from other memory solutions by providing local, persistent PostgreSQL storage, delivery programs with evidence‑gated checkpoints, per‑agent HMAC audit trails, identity isolation, skill admission gating, tamper‑evident session replay, a work‑stealing queue for parallel sessions, and offline operation with $0 idle cost. Competing services either store data in the cloud, offer only vector RAG without state management, or lack audit and delivery features.

Question 10

What's new (v0.46.1)?

Accepted Answer

Version 0.46.1 adds Temporal retrieval round 2 and the delivery‑tool kit. Compound temporal questions are decomposed into event clauses with rank fusion; temporal searches render a chronological Timeline block and stale results carry age warnings. The new zcprogram adds multi‑phase delivery programs with evidence‑gated, auto‑checkpointing phase closure, verified by an orchestrator handoff test. Additional improvements include a watchdog for the embed lane, budgeted self‑healing, and many enhancements listed in the changelog.

Question 11

What is SecureContext?

Accepted Answer

SecureContext is an MCP plugin that provides persistent memory across Claude Code restarts, per‑file semantic summaries, temporal search, a delivery layer for long‑running projects, and a cryptographic audit trail. It stores decisions, gotchas, and project state in PostgreSQL, generates searchable checkpoints, and records every agent action with tamper‑evident logs.

Огляд

Ключові функції

Ціни

Кейси використання

Повісна підготовка фахівця

Контроль якості аналізу даних

Плюси і мінуси

Плюси

Мінуси

Відгуки

Питання

Постав питання

Альтернативи Мовні здібності