multi-window

A context system for keeping work state across separate Claude Code sessions. It saves progress at checkpoints so a later session can resume after an interruption.

In plain words
What is it for?
Use it to checkpoint multi-step work, resume unfinished phases, and continue a workflow in a new session.
Why use it?
It prevents long-running work from losing its place when a session ends or is interrupted.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/thibautmelen/agentic-ai-systems/multi-window
Clone the repo
git clone --depth 1 https://github.com/ThibautMelen/agentic-ai-systems
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,623 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.01623
Opus 5 $0.00000 $0.00812
Sonnet 5 $0.00000 $0.00325
Haiku 4.5 $0.00000 $0.00162

Measured yesterday against content hash 5b378c76fcfc, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

multi-window scanned grade A with 0 findings against 26 rules in 11 categories โ€” prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency โ€” measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/multi-window.md ยท 225 lines

How it starts

The opening of the file, as written. The whole thing โ€” 225 lines โ€” stays where its author put it; the contents beside it link to each section on GitHub.

๐Ÿ  Home โ€บ Autonomous โ€บ ๐Ÿ–ฅ๏ธ Multi-Window Context

โ† Autonomous Agent โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ—โ”โ” Autonomous โ†’


๐Ÿ–ฅ๏ธ Multi-Window Context

TL;DR: State persistence across Claude Code sessions. Checkpoint your progress, resume from interruptions. Essential for long-running workflows.


Diagram

%%{init: {'theme': 'base', 'themeVariables': {'lineColor': '#64748b'}}}%%
flowchart TB
    classDef checkpoint fill:#f59e0b,stroke:#d97706,stroke-width:2px,color:#ffffff
    classDef state fill:#10b981,stroke:#059669,stroke-width:2px,color:#ffffff

    subgraph Session1["Session 1"]
        S1P1["๐Ÿ—๏ธ Phase 1"] --> S1CP["๐Ÿ–ฅ๏ธ Checkpoint"]:::checkpoint
        S1CP --> S1P2["๐Ÿ”— Phase 2"]
        S1P2 --> INTERRUPT["โŒ Interrupt"]
    end

    subgraph Session2["Session 2 (Resume)"]
        RESUME["๐Ÿ”„ Resume"] --> S2P2["Continue Phase 2"]
        S2P2 --> S2P3["๐Ÿ“ Phase 3"]
        S2P3 --> DONE["โœ… Complete"]
    end

    S1CP -.->|๐Ÿ’พ State saved| STATE[("๐Ÿ’พ State Store")]:::state
    STATE -.->|๐Ÿ’พ State loaded| RESUME

    classDef session1Box fill:#fef2f2,stroke:#ef4444,stroke-width:2px,color:#991b1b
    classDef session2Box fill:#ecfdf5,stroke:#10b981,stroke-width:2px,color:#065f46

    Session1:::session1Box
    Session2:::session2Box

When to Use

  • Large-scale generation (1000+ files)
  • Long research tasks
  • Multi-day workflows
  • Error recovery needs
  • Context window limitations

Implementation

Session 1:
  1. Work on Phase 1
  2. Save checkpoint (progress, decisions, context)
  3. Continue until context limit or interruption

Session 2:
  1. Load checkpoint
  2. Resume from saved state
  3. Continue execution

Flow Example

๐Ÿ–ฅ๏ธ MULTI-WINDOW CONTEXT
Session 1: ๐Ÿ™‹โ€โ™€๏ธ๐Ÿ“ฅ โ”€โ”€โ–บ ๐Ÿ”๐Ÿ“‹ โ”€โ”€โ–บ ๐Ÿ”โšก โ”€โ”€โ–บ ๐Ÿ–ฅ๏ธ๐Ÿ’พ โ”€โ”€โ–บ [Context Limit]
                                      โ†“
Session 2: ๐Ÿ–ฅ๏ธ๐Ÿ’พ โ”€โ”€โ–บ ๐Ÿ”โšก โ”€โ”€โ–บ ๐Ÿ”๐Ÿ‘€ โ”€โ”€โ–บ ๐Ÿ”๐Ÿ’ญ โ”€โ”€โ–บ ๐Ÿ’โ€โ™€๏ธ๐Ÿ“ค

Checkpointing for Long Workflows

Read the full file on GitHub ยท 225 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen ยท 225 lines ยท 0 tokens per session scan A 5b378c76fcfc

Subscribe to this mod's changes

multi-window is an agent published in the GitHub repository ThibautMelen/agentic-ai-systems (301 stars, last pushed 1mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,623 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other agents, from other repositories

generate_agent

Generates a customized agent based on user-defined parameters.

ai-driven-dev/framework ยท 14 tokens

external-system-integration-expert

ไฝ ่ดŸ่ดฃๆŠŠๅฝ“ๅ‰้กน็›ฎไธŽๅค–้ƒจ APIใ€API ็ฝ‘ๅ…ณๅŠไธšๅŠก็ณป็ปŸๅฎ‰ๅ…จๅœฐ่ฟžๆŽฅ่ตทๆฅ๏ผš่ฏ†ๅˆซ้›†ๆˆ่พน็•Œใ€ๆ•ด็†ๆŽฅๅฃไธŽ็Žฏๅขƒๅทฎๅผ‚ใ€้ชŒ่ฏ่ฏทๆฑ‚ๅ’Œๅ“ๅบ”ใ€ๅฎšไฝ่ฎค่ฏๆˆ–ๆ•ฐๆฎๅฅ‘็บฆ้—ฎ้ข˜ใ€‚.

agents-universe/agents-universe ยท 33 tokens

ba-designer

Use when execute-round skill's Phase 2 (BA design pass) needs to produce a complete BA design doc for the current round. Generates D-1..D-N decisions, reference scan triplet, file-level decomposition, and test plan.

Arch1eSUN/Arcgentic ยท 53 tokens

vc-innovate-agent

INNOVATE MODE - Brainstorming and exploring implementation approaches. Discusses possibilities without making decisions. Use after research is complete.

withkynam/vibecode-pro-max-kit ยท 32 tokens

design-reviewer

Design lead + expert design critic. Two modes: Mode A โ€” authors the project's root DESIGN.md (design identity) at project start. Mode B โ€” reviews built UI against DESIGN.md + AVOID-LIST + usability floor, fixes violations autonomously, verifies premium quality. Delegate when: a UI project has no DESIGN.md yet, UIโ€ฆ

wasintoh/toh-framework ยท 85 tokens

Audit

Deep security + performance audit of a specific diff. Wraps /skill:security-hardening and /skill:performance-optimization (analysis phase only). Use when a change touches auth, untrusted input, secrets, webhooks, PII, or a latency/throughput budget โ€” a focused, read-only risk pass that returns findings the parentโ€ฆ

BlackBeltTechnology/pi-agent-dashboard ยท 98 tokens