nexus-prime 7.9.36 → 7.9.38

package/README.md

@@ -38,14 +38,15 @@
   <a href="https://www.producthunt.com/products/nexus-prime?embed=true&utm_source=badge-featured&utm_medium=badge&utm_campaign=badge-nexus-prime" target="_blank" rel="noopener noreferrer"><img alt="Nexus-Prime — Product Hunt" width="250" height="54" src="https://api.producthunt.com/widgets/embed-image/v1/featured.svg?post_id=1096831&theme=dark&t=1773345508816"></a>
 </div>
 
-> **Stop restarting your coding agents from zero.**
+> **Give every coding agent the same memory, budget, and proof trail.**
 >
-> Nexus Prime gives Claude Code, Codex, Cursor, Windsurf, OpenCode, Aider, and the rest of your local agent stack one shared memory, one proof trail, and one dashboard for what actually happened.
+> Nexus Prime is the local-first operating layer for Claude Code, Codex, Cursor, Windsurf, OpenCode, Aider, and the rest of your agent stack. It helps them remember prior work, route context before they burn tokens, coordinate across sessions, and show you exactly what happened before a change ships.
 
 ---
 
 ## Contents
 
+- [Why teams install it](#why-teams-install-it)
 - [The problem](#the-problem)
 - [What changes when you install](#what-changes-when-you-install)
 - [Quick start](#quick-start)
@@ -54,7 +55,7 @@
 - [See it in action](#see-it-in-action)
 - [Competitive position](#competitive-position)
 - [Works with](#works-with)
-- [What users see](#what-users-see)
+- [What the dashboard proves](#what-the-dashboard-proves)
 - [Who it's for](#who-its-for)
 - [Pricing](#pricing)
 - [FAQ](#faq)
@@ -62,41 +63,38 @@
 - [Community & support](#community--support)
 - [License](#license)
 
----
-
-## Docs and comparison
-
-- Website: https://nexus-prime.cfd/
-- Comparison: https://nexus-prime.cfd/comparison.html
-- Knowledge base: https://nexus-prime.cfd/knowledge-base.html
-
-Agents should start with `nexus_session_bootstrap`, then route the raw request through `nexus_orchestrate`. Auto-bootstrap runs on first non-bootstrap call for compatible clients, but the explicit bootstrap-first path remains the safest integration contract.
-
----
-
 ## Why teams install it
 
 Nexus Prime is the local-first control plane for serious AI-assisted coding.
 
-- **Shared memory for every agent.** Your tools stop acting like strangers and start carrying useful context across sessions.
-- **Lower token waste.** Nexus routes agents toward the files and facts that matter instead of rereading the repo every turn.
-- **Runtime truth you can inspect.** The dashboard shows runs, memory, token savings, and agent activity instead of hiding the messy parts.
-- **Safer multi-file work.** Planning, verification, and handoff state stay attached to the repo, so interrupted work is easier to resume and review.
-- **Local by default.** Your code and agent memory stay on your machine unless you explicitly opt into something else.
+- **Memory that compounds.** Your agents stop acting like strangers and start carrying useful context across sessions, tools, and repos.
+- **Token spend you can manage.** Nexus routes agents toward the files and facts that matter before they reread the whole repo.
+- **Runtime truth for real work.** The dashboard shows runs, memory, token savings, worker activity, and verification state instead of hiding the messy parts.
+- **Safer multi-file execution.** Planning, review, and handoff state stay attached to the repo, so interrupted work is easier to resume and audit.
+- **A product your team can trust.** Local-first storage, opt-in telemetry, 15-day trial enforcement, and license-gated runtime work are built into the product surface.
+
+Install Nexus Prime when:
+
+- You use more than one coding agent and keep repeating the same context.
+- Your agent bill is mostly input tokens and repo rereads.
+- You need proof of what the agent planned, touched, skipped, and verified.
+- You want a local control plane before putting AI-assisted coding in front of a team.
 
 ---
 
 ## The problem
 
-Every coding agent you use starts every session cold.
+Every coding agent starts cold.
 
-It has no idea what you worked on yesterday. It has no idea what your teammate's Cursor session just finished. It re-reads your repo from scratch, every single turn, for every single question — because that's the only thing it knows how to do.
+It has no idea what you worked on yesterday. It has no idea what your teammate's Cursor session just finished. It does not know which files already mattered, which decisions were accepted, which paths failed, or which preferences should carry forward.
 
-One tracked session was measured at **100,000,000 tokens**. Ninety-nine point four percent of them were input. **166 tokens read for every 1 written.** Your agent isn't thinking. It's re-orienting. And you're paying for that re-orientation over and over.
+So it re-reads your repo from scratch, every turn, for every question, because cold context is the only context it has.
 
-This is **agent amnesia** — and it's the quiet tax on every hour you spend with AI-assisted coding.
+One tracked session was measured at **100,000,000 tokens**. Ninety-nine point four percent of them were input. **166 tokens read for every 1 written.** That is not deep reasoning. That is repeated re-orientation.
 
-**Nexus Prime turns that cold start into a remembered workspace.**
+This is **agent amnesia**: the quiet tax on every serious AI-assisted engineering workflow.
+
+**Nexus Prime turns cold starts into a remembered workspace.**
 
 ---
 
@@ -105,12 +103,12 @@ This is **agent amnesia** — and it's the quiet tax on every hour you spend wit
 | Before Nexus Prime | After Nexus Prime |
 | :--- | :--- |
 | Every session starts cold | Your agents pick up where they (or any other agent) left off |
-| You pick files manually and pray they fit the context | Your agent reads only what matters — automatically |
-| Memory grows but nothing learns | What was true yesterday is still true today |
-| Multi-file edits sometimes leave the codebase in a half-broken state | Every change is reviewed and safe before it lands |
-| Your Cursor and your Claude Code sessions are two strangers | They share the same memory, in real time |
-| Token bills climb and you can't see why | A live dashboard shows exactly where every token went |
-| Five-figure monthly LLM spend and no lever to pull | **Users typically cut input tokens by 60–90%** |
+| You pick files manually and hope they fit | Nexus routes context before the agent starts reading |
+| Memory grows in random chat threads | Reusable project memory lives with the workspace |
+| Multi-file edits can die halfway through | Plans, reviews, and handoffs remain inspectable |
+| Cursor, Claude Code, and Codex act separately | They share a common local control plane |
+| Token bills climb and you cannot explain why | The dashboard shows where context went and what was saved |
+| You have no leverage over input-token waste | **Users typically cut input tokens by 60–90%** |
 
 ---
 
@@ -153,7 +151,7 @@ Then visit `http://localhost:3377`.
 
 ### 4. Use your agent like normal
 
-That's it. Memory, budgeting, and safe execution are already on.
+That's it. Memory, routing, budgeting, and proof capture are already on.
 
 </td>
 </tr>
@@ -161,6 +159,18 @@ That's it. Memory, budgeting, and safe execution are already on.
 
 Prefer one-liner installs? `curl -fsSL https://nexus-prime.cfd/install.sh | bash`
 
+You get a 15-day local trial. After the trial, runtime work requires an active license.
+
+---
+
+## Docs and comparison
+
+- Website: https://nexus-prime.cfd/
+- Comparison: https://nexus-prime.cfd/comparison.html
+- Knowledge base: https://nexus-prime.cfd/knowledge-base.html
+
+Agents should start with `nexus_session_bootstrap`, then route the raw request through `nexus_orchestrate`. Auto-bootstrap runs on first non-bootstrap call for compatible clients, but the explicit bootstrap-first path remains the safest integration contract.
+
 ---
 
 ## Features
@@ -183,9 +193,9 @@ Your agent reads what matters — nothing else. Users typically see **60–90% f
 </td>
 <td width="33%" valign="top">
 
-### 🛡️ Safer coding sessions
+### 🛡️ Reviewable coding sessions
 
-Every multi-file edit is reviewed before it lands. No more half-broken commits because the agent got interrupted.
+Plans, file choices, worker activity, and verification state stay visible while the agent works.
 
 </td>
 </tr>
@@ -255,6 +265,14 @@ Install, run `setup all`, done. The wizard detects every coding agent on your ma
   <sub><b>The receipts</b> — every decision is traced, so you can see <i>why</i> your agent did what it did.</sub>
 </div>
 
+<br>
+
+<div align="center">
+  <img src="https://nexus-prime.cfd/assets/screenshots/dashboard_runtime_sequence.png" alt="Nexus Prime runtime sequence — connected agents, memory graph, and guarded operator actions" width="1024">
+  <br>
+  <sub><b>The runtime</b> — connected agents, guarded actions, memory state, and work context in one local view.</sub>
+</div>
+
 ---
 
 ## Works with
@@ -277,25 +295,36 @@ Or just wire everything: **`nexus-prime setup all`**
 
 ---
 
-## What users see
+## What the dashboard proves
 
-Numbers from real users running real workflows on their own machines:
+Your local dashboard turns the sales promise into inspectable evidence:
 
-| Metric | Typical before | Typical after | What it means |
-|--------|---------------:|--------------:|:--------------|
-| Input tokens per session | 10M – 100M | 1M – 10M | **60–90% lower bills** |
-| Cold-start context time | 30s – 2 min | under a second | Agents feel instant |
-| Memory across sessions | None | Full | Nothing gets forgotten |
-| Multi-agent state sharing | None | Real-time | Your tools stop fighting each other |
-| Half-broken multi-file edits | Common | Rare | Merge with confidence |
+| Surface | What you can inspect | Why it matters |
+|:--------|:---------------------|:---------------|
+| Token telemetry | Input, forwarded, saved, and compressed tokens | Proves whether Nexus is actually saving context |
+| Memory lattice | Accepted memories, recalled facts, and learning packets | Shows what the system learned and reused |
+| Runtime activity | Runs, workers, routes, and verification state | Makes agent work observable instead of mystical |
+| Context log | Selected files, changed files, and route rationale | Explains why the agent read what it read |
+| License status | Trial, plan, expiry, and activation state | Keeps commercial use explicit |
 
-No benchmarks from a press release — this is what the dashboard shows you, on day one.
+Typical users see **60–90% lower input-token use** once Nexus has enough project memory to route context well. The important part: your own dashboard shows your own numbers.
 
 ---
 
 ## Competitive position
 
-Nexus Prime is a local-first control plane for coding agents: persistent memory, workspace-scoped runtime state, dashboard observability, and MCP-native orchestration in one install. The public comparison page tracks where it differs from other agent frameworks and developer orchestrators: https://nexus-prime.cfd/comparison.html
+Most agent tools focus on a single chat, editor, or framework. Nexus Prime focuses on the workspace underneath them.
+
+It gives your agent stack:
+
+- persistent memory
+- workspace-scoped runtime state
+- token routing and compression telemetry
+- dashboard observability
+- MCP-native orchestration
+- license-aware local execution
+
+The public comparison page tracks where it differs from other agent frameworks and developer orchestrators: https://nexus-prime.cfd/comparison.html
 
 ---
 
@@ -309,7 +338,7 @@ Nexus Prime is for people who take AI-assisted coding seriously:
 - **Power users** who've hit every context-window wall and want a real fix
 - **Anyone** who's paid a four- or five-figure monthly LLM bill and felt it
 
-If any of that sounds like you — keep reading.
+If you only use an agent once a month, you probably do not need Nexus Prime yet. If AI is already part of how you ship, Nexus gives that workflow memory, controls, and receipts.
 
 ---
 
@@ -317,7 +346,7 @@ If any of that sounds like you — keep reading.
 
 <div align="center">
 
-**15-day local trial. Activate or upgrade before expiry; runtime work locks after day 15.**
+**Start with a 15-day local trial. Activate or upgrade before expiry; runtime work locks after day 15.**
 
 </div>
 
@@ -371,7 +400,7 @@ nexus-prime license status          # check your plan
 nexus-prime license activate <key>  # activate the key shared by the team
 ```
 
-→ **[nexus-prime.cfd/account](https://nexus-prime.cfd/account)** · **[Request by email](mailto:adarsh@nexus-prime.cfd?subject=Nexus%20Prime%20license%20request)**
+Start here: **[nexus-prime.cfd/account](https://nexus-prime.cfd/account)** · **[Request by email](mailto:adarsh@nexus-prime.cfd?subject=Nexus%20Prime%20license%20request)**
 
 ---
 
@@ -446,7 +475,6 @@ Nexus Prime was designed privacy-first, because the code on your machine is your
 - 🏠 **100% local.** All data lives on your disk, in your home directory.
 - 🚫 **No cloud sync.** Your code, your memory, your logs — never uploaded.
 - 🔐 **Bounded local trial.** You can run locally for 15 days without signing up; a license is required after the trial expires.
-- 🧱 **Binary-first release path.** The current NPM package remains intact, and a prepared shim lane can turn NPM into a signed-binary downloader once binary artifacts are published and verified.
 - 📊 **Telemetry is opt-in.** Off by default. When on, only aggregate, anonymous event counts are sent — never source code, never memory contents.
 - 🗑️ **Uninstall is clean.** Everything lives in one directory you can delete.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "nexus-prime",
-  "version": "7.9.36",
+  "version": "7.9.38",
   "description": "Local-first MCP control plane for coding agents with bootstrap-orchestrate execution, memory fabric, token budgeting, and worktree-backed swarms (binary installer shim)",
   "type": "module",
   "bin": {