# Sirro

> Sirro is a library of reusable work for AI agents. Save work once across coding tools, and let Playbooks find the cheapest model that does a task. Personal-agent Playbooks are in development.

Sirro keeps reusable work available across agents, projects, and teams. Coding agents save and retrieve named assets, Packs and Playbooks. A Playbook is a task written down by a strong model so a cheaper one can follow it, with a run log that names the cheapest model that passes. Company workspaces support shared knowledge and onboarding. Personal and company Playbooks for everyday work are in development.

## Canonical pages

- [Home](https://sirro.dev/): Product overview for Sirro, a library of reusable work for coding and personal agents.
- [Sirro](https://sirro.dev/sirro-code): Persistent access for coding agents to reusable components, patterns, skills, rules, and development knowledge.
- [Sirro Whiteboard](https://sirro.dev/whiteboard): A shared, persistent environment where AI agents coordinate tasks, context, rules, files, and ongoing work.
- [Sirro Ops](https://sirro.dev/sirro-ops): An operating environment for the business. Sirro Ops is coming soon.
- [About Sirro](https://sirro.dev/about): Sirro keeps reusable work and proven methods available across agents, projects, and teams.
- [What is Sirro](https://sirro.dev/what-is-sirro): Sirro for coding agents, enterprise teams, and personal workflows, including Playbooks and the planned personal-agent library.
- [FAQ](https://sirro.dev/faq): Answers about coding-agent libraries, Packs, Playbooks, sharing, and access.
- [Use cases](https://sirro.dev/use-cases): Shared components, saved prompts, and a team library for people and their agents.
- [Shared components across repositories](https://sirro.dev/use-cases/shared-components): Save a UI component once in Sirro and retrieve it by name in another repository.
- [Prompts and agent instructions](https://sirro.dev/use-cases/prompts): Save prompts worth repeating so any connected agent can retrieve them by name.
- [A team library for people and agents](https://sirro.dev/use-cases/team-library): One team workspace so people and their allowed agents share the same asset names.
- [How it works](https://sirro.dev/how-it-works): Workflow for saving reusable assets once and retrieving them by name in any project.
- [Integrations](https://sirro.dev/integrations): Connect Sirro to Cursor, Claude Code, Codex, and other AI coding agents. Sign in and Allow.
- [For enterprise](https://sirro.dev/enterprise): Company onboarding, reusable knowledge, workspace access, and planned company Playbooks.
- [Agent onboarding](https://sirro.dev/docs/agent-onboarding): Create and maintain a Pack of working knowledge for a new teammate and their agent.
- [How work improves](https://sirro.dev/docs/improvement): How saved work and Playbooks improve through outcomes, evidence, versions, and review.
- [Compute and reuse](https://sirro.dev/docs/compute-efficiency): How retrieval replaces repeated generation and how to measure total task cost.
- [Documentation](https://sirro.dev/docs): Connect coding agents, run Playbooks on the cheapest model that passes, and read about planned personal-agent Playbooks.
- [Playbooks](https://sirro.dev/docs/playbooks): A task written down by a strong model so a cheaper one can follow it, with the record of which models pass it and the cheapest one that does.
- [Personal agents and Playbooks](https://sirro.dev/docs/personal-agents): The planned personal agent harness: Playbooks, confirmed work facts, versions, proof, privacy, and cross-agent reuse.
- [Developer](https://sirro.dev/developer): Connect agents, use Packs, and share reusable work across your Sirro workspaces.
- [Setup](https://sirro.dev/developer/setup): Connect an agent, Allow it in the browser, then save and reuse by name.
- [Tips](https://sirro.dev/developer/tips): Guides to onboard agents, use one library across tools, and share Packs with your Team.
- [Agent onboarding](https://sirro.dev/developer/tips/agent-onboarding): Give a teammate's agent the docs and workflows it needs with one Pack.
- [One library, every agent](https://sirro.dev/developer/tips/one-library-every-agent): Save in Claude Code, then use the same work in Codex or Cursor.
- [Share the good stuff with your Team](https://sirro.dev/developer/tips/share-with-your-team): Add a Pack to your Team workspace so coworkers and their agents can use it.
- [Cursor memory for reusable coding assets](https://sirro.dev/for/cursor-memory): Use Sirro as shared memory for Cursor across coding projects and repositories.
- [Claude Code memory across coding projects](https://sirro.dev/for/claude-code-memory): Save named assets with Claude Code and retrieve them in later sessions or repositories.
- [Codex memory for reusable code and components](https://sirro.dev/for/codex-memory): Connect Codex to a Sirro library and reuse saved code across repositories.
- [AI agent memory for reusable coding work](https://sirro.dev/for/ai-agent-memory): A deliberate asset library that connected AI coding agents can use across projects.
- [Shared context for coding agents across projects](https://sirro.dev/for/shared-context-coding-agents): Keep named code assets and links available to permitted coding agents outside chat sessions.
- [MCP memory server for reusable coding assets](https://sirro.dev/for/mcp-memory-server): Save and retrieve named coding assets across projects through the Sirro MCP server.
- [Articles](https://sirro.dev/articles): Writing on memory, the environments, and how an agent reads and writes Sirro.
- [From generation to composition](https://sirro.dev/articles/from-generation-to-composition): The work your team has already approved can become the starting point for the next project.
- [Memory designed for coding agents](https://sirro.dev/articles/memory-designed-for-coding-agents): Useful memory starts with the right result at the moment an agent needs it.
- [How Sirro fits into a coding workflow](https://sirro.dev/articles/how-environments-fit): Your project, coding agent, and reusable library each have a clear place.
- [How memory holds across agents and people](https://sirro.dev/articles/how-memory-holds): Give finished work a name and a place where your team and its agents can find it again.
- [Better models need better harnesses](https://sirro.dev/articles/better-models-need-better-harnesses): The right context, tools, and saved work help coding agents make more of their capability.
- [Marketplace](https://sirro.dev/marketplace): Reusable UI, skills, workflows, agent setups, architecture and prompts, published by people and ready for your agents.
- [News](https://sirro.dev/news): Sirro product milestones and company updates.
- [Security](https://sirro.dev/security): Where Sirro keeps your data, who processes it, how access works, how long things are kept, and how to reach us.
- [Privacy Policy](https://sirro.dev/privacy): How Sirro collects, uses, and protects your data.
- [Terms of Service](https://sirro.dev/terms): Terms governing use of the Sirro service.

## AI agent and team collaboration

- Connect multiple AI coding agents (Cursor, Claude Code, Codex, and other coding tools) to one Sirro library
- Share saved components, prompts, patterns, and links across teammates and their agents
- Reuse assets by name in any repository or project without copy-paste
- Keep one source of truth for reusable work across your team's coding stack

## What Sirro does

- Stores reusable dev assets outside any Git repository
- Lets AI agents save and retrieve assets by name
- Supports React and UI components, animations and motion patterns, page sections and layouts, prompts and agent instructions, code patterns and snippets, reference links and documentation
- Works with Cursor, Claude Code, Codex, Antigravity, GitHub Copilot, Devin, and other coding tools
- Offers personal and team libraries with versioning and search

## Enterprise and onboarding

Sirro company workspaces let owners invite teammates and connect agents to a shared library. Packs collect docs, examples, and workflows for a role or project. A new teammate asks their agent to use the onboarding Pack by name. Workspace membership controls access; content already copied into a project cannot be retracted by removing access.

## Playbooks

A Playbook is a task a strong model did once, written down so a cheaper model can follow it: numbered steps, every judgement call as an explicit rule, worked examples with the exact input and correct output, a done check, and pitfalls. The user says "save this as a playbook" and the agent saves it by name; any connected agent can then run it. Every run records the agent, the model and whether it passed. A run counts only when the user confirmed it or Sirro graded the agent's answers to the examples; the agent's own report never counts. The library orders the models that tried cheapest first and names the cheapest model that passes. When a cheaper model fails, a strong model tightens the steps into a new version, and runs stay attached to the version they followed. Coding-agent Playbooks are live.

## Improvement and personal agents

Saved coding assets and Packs are updated and retrieved across projects and tools. Sirro is also developing personal and company Playbooks for recurring everyday work, with confirmed work facts, outcome records, versions, and restoration. The planned stages are draft, used, and proven. Proven requires a person-confirmed result or checkable evidence. Company Playbook changes require owner approval. Personal sharing is an explicit choice. These changes improve reusable methods, not the underlying model weights.

## Compute efficiency

Retrieving a saved implementation or method replaces repeated generation and reconstruction with retrieval and adaptation. Retrieval, context reading, adaptation, validation, and human review also cost resources. Evaluate complete tasks against the same quality criteria. The net effect depends on the work avoided and the cost of reuse. Quantitative results require measurements on comparable tasks.

## Agent tools

- `sirro_save`: Save code or a link to your Sirro library.
- `sirro_get`: Retrieve a saved asset by name.
- `sirro_list`: List saved asset and Pack names.
- `sirro_update`: Update metadata for a saved asset.
- `sirro_compose`: Fetch named assets or a Pack in one request and write them into the project.
- `sirro_save_bundle`: Save a Pack of related or ordered asset names.
- `sirro_open`: Open the Sirro library in the browser.

## Who it is for

Developers reusing work across projects, enterprises onboarding people and agents into company workflows, and people seeking repeatable methods for personal agents. Personal and company Playbook workflows are in development.

## How it works

1. Create a Sirro account and library
2. Connect your agent from the Sirro library, then sign in and Allow in the browser
3. Save an asset once from any project
4. Reuse it by name in any future project with any connected agent on your team
5. After a strong model does a task well, say "save this as a playbook", then run it with a cheaper model and confirm the result

## Example agent prompts

"Use Animated Button from Sirro" in a new repo pulls the saved component into the current project.

"Run the invoice-classification playbook from Sirro on this" fetches the steps, opens a run, and records which model ran it and whether it passed.

## Contact

hello@sirro.dev

## Optional

- [Extended machine-readable overview](https://sirro.dev/llms-full.txt)
- [XML sitemap](https://sirro.dev/sitemap.xml)
- [Robots](https://sirro.dev/robots.txt)