claude-overnight 1.16.4 → 1.16.5
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +17 -20
- package/package.json +2 -2
package/README.md
CHANGED
|
@@ -1,10 +1,12 @@
|
|
|
1
1
|
# claude-overnight
|
|
2
2
|
|
|
3
|
-
**
|
|
3
|
+
**A background lane for your Claude Max plan.** Runs a capped swarm of Claude Agent SDK sessions in isolated git worktrees — stops at a usage cap you set, so your interactive Claude Code always has headroom. Rate-limited? It waits. Crash? It resumes with full context.
|
|
4
4
|
|
|
5
|
-
|
|
5
|
+
Your Max plan rate limits eat interactive coding time. One deep refactor and the 5-hour window is gone before lunch. `claude-overnight` runs background agent sessions up to the percentage cap you pick (90% is typical), leaving the rest free for your own Claude Code session. Hand it an objective and a session budget, walk away, review the diff when the run ends.
|
|
6
6
|
|
|
7
|
-
|
|
7
|
+
Isolated by default. Every agent runs in its own git worktree on its own branch, so a misbehaving agent can't trash your working tree. You choose what agents can do before the run starts — no surprise escalation mid-flight. Unmerged branches are preserved for manual review, never discarded. Built on the [Claude Agent SDK](https://www.npmjs.com/package/@anthropic-ai/claude-agent-sdk) — not a Claude Code replacement, but a background lane that runs alongside it.
|
|
8
|
+
|
|
9
|
+
Different shape from hosted agent harnesses like [Claude Managed Agents](https://platform.claude.com/docs/en/managed-agents/overview): instead of one agent in one cloud container billed separately, you get many parallel sessions on your own machine, in your real repo, against your own Max plan (or API key). Works with Claude Opus, Sonnet, and Haiku — or pair an Anthropic planner with a cheaper executor on Qwen, OpenRouter, or any Anthropic-compatible endpoint.
|
|
8
10
|
|
|
9
11
|
## Install
|
|
10
12
|
|
|
@@ -53,27 +55,22 @@ claude-overnight
|
|
|
53
55
|
|
|
54
56
|
◆ Thinking: 5 agents exploring... ← architects analyze your codebase
|
|
55
57
|
◆ Orchestrating plan... ← synthesizes 50 concrete tasks
|
|
56
|
-
◆ Wave 1 · 50 tasks · $4.20 spent ←
|
|
58
|
+
◆ Wave 1 · 50 tasks · $4.20 spent ← runs unattended from here
|
|
57
59
|
↑ 1.2M in ↓ 340K out $4.20 / $4.24 total
|
|
58
60
|
◆ Assessing... how close to amazing?
|
|
59
61
|
◆ Wave 2 · 30 tasks · $18.50 spent ← improvements from assessment
|
|
60
62
|
◆ Reflection: 2 agents reviewing ← deep quality audit
|
|
61
63
|
◆ Wave 3 · 20 tasks · $31.00 spent ← fixes from review findings
|
|
62
|
-
◆ Assessing... ✓
|
|
64
|
+
◆ Assessing... ✓ Done
|
|
63
65
|
```
|
|
64
66
|
|
|
65
|
-
You interact once (objective, budget, model, review themes), then
|
|
66
|
-
|
|
67
|
-
## How is this different?
|
|
68
|
-
|
|
69
|
-
Claude already has several ways to run agents. `claude-overnight` fills a specific niche:
|
|
67
|
+
You interact once (objective, budget, model, review themes), then the rest runs unattended — thinking, planning, executing, reflecting, steering. Rate-limited? It waits and retries. Crash? Resume where you left off. Capped at usage limit? Pick up next time with full context preserved.
|
|
70
68
|
|
|
71
|
-
|
|
72
|
-
- **[Claude Managed Agents](https://platform.claude.com/docs/en/managed-agents/overview)** — a hosted single-session agent harness. One agent, one cloud container, stateful conversation. `claude-overnight` is a local multi-session *orchestrator* built on the [Claude Agent SDK](https://www.npmjs.com/package/@anthropic-ai/claude-agent-sdk): many parallel sessions on your machine, in your real repo, with spend caps and multi-day crash-safe resume.
|
|
73
|
-
- **[Claude Agent SDK](https://www.npmjs.com/package/@anthropic-ai/claude-agent-sdk)** — primitives for building your own agent. `claude-overnight` is one specific thing built on top of it: an overnight swarm orchestrator you didn't have to write.
|
|
74
|
-
- **IDE copilots (Cursor, Copilot, Cline, etc.)** — synchronous assistants that complete while you're at the keyboard. `claude-overnight` is asynchronous: you hand off an objective and a budget, close the laptop lid, and review a branch in the morning.
|
|
69
|
+
## How it differs
|
|
75
70
|
|
|
76
|
-
|
|
71
|
+
- vs **Claude Code**: many agents, no driver, capped so your Claude Code session keeps its headroom
|
|
72
|
+
- vs **[Managed Agents](https://platform.claude.com/docs/en/managed-agents/overview)**: on your machine, against your Max plan, in your real git history — not a cloud container billed separately
|
|
73
|
+
- vs **Cursor / Copilot / Cline**: asynchronous, off the keyboard
|
|
77
74
|
|
|
78
75
|
## Use cases
|
|
79
76
|
|
|
@@ -86,17 +83,17 @@ If you want to hand an objective and a spend cap to Claude and wake up to shippe
|
|
|
86
83
|
- **Quality audits** — reflection waves surface architectural issues and code smells.
|
|
87
84
|
- **Long research runs** — architect sessions explore a large codebase before any code lands.
|
|
88
85
|
|
|
89
|
-
Typical shape: one objective + a $20–$200 spend cap +
|
|
86
|
+
Typical shape: one objective + a $20–$200 spend cap + walk away.
|
|
90
87
|
|
|
91
88
|
## How it works
|
|
92
89
|
|
|
93
|
-
### 1. Thinking
|
|
90
|
+
### 1. Thinking phase — parallel architect sessions
|
|
94
91
|
|
|
95
92
|
For budgets > 15, the tool launches **architect agents** that explore your codebase before any code is written. Each one gets a different research angle (architecture, data models, APIs, testing, etc.) and writes a structured design document. The number scales with budget: 5 for budget=50, 10 for budget=2000.
|
|
96
93
|
|
|
97
94
|
### 2. Task orchestration
|
|
98
95
|
|
|
99
|
-
An orchestrator session reads all design documents and synthesizes concrete execution tasks — grounded in real files and patterns the architects found.
|
|
96
|
+
An orchestrator session reads all design documents and synthesizes concrete execution tasks — grounded in real files and patterns the architects found. The task plan is also written to a file for resilience — if orchestration is interrupted, partial results survive.
|
|
100
97
|
|
|
101
98
|
### 3. Parallel execution waves
|
|
102
99
|
|
|
@@ -110,7 +107,7 @@ After each wave, steering assesses: "how good is this?" — not "what's missing?
|
|
|
110
107
|
|
|
111
108
|
### 4. Goal refinement and steering
|
|
112
109
|
|
|
113
|
-
The tool starts with your broad objective but
|
|
110
|
+
The tool starts with your broad objective but refines its definition of quality as it learns your codebase. Steering updates the goal after each wave. Late waves are informed by early discoveries.
|
|
114
111
|
|
|
115
112
|
### 5. Three-layer context memory
|
|
116
113
|
|
|
@@ -118,7 +115,7 @@ Long runs stay sharp because steering maintains three layers of memory:
|
|
|
118
115
|
|
|
119
116
|
- **Status** — a living project snapshot, updated every wave. Compressed, never truncated.
|
|
120
117
|
- **Milestones** — strategic snapshots archived every ~5 waves. Long-term memory.
|
|
121
|
-
- **Goal** — the evolving north star. What
|
|
118
|
+
- **Goal** — the evolving north star. What quality means for this codebase.
|
|
122
119
|
|
|
123
120
|
## Run history, resume, and knowledge carryforward
|
|
124
121
|
|
package/package.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "claude-overnight",
|
|
3
|
-
"version": "1.16.
|
|
4
|
-
"description": "
|
|
3
|
+
"version": "1.16.5",
|
|
4
|
+
"description": "Background lane for your Claude Max plan. Parallel Claude Agent SDK sessions in git worktrees with a usage cap that reserves headroom for your interactive Claude Code. Crash-safe resume. Opus/Sonnet/Haiku + Qwen/OpenRouter.",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"bin": {
|
|
7
7
|
"claude-overnight": "dist/bin.js"
|