One memory.Every session.Across every agentic assistant.
Spooling indexes every AI coding session across your tools into one searchable memory. Spooling Cloud makes that memory shared across your team.
The OSS engine. The Cloud SaaS. Same install path.
Every developer installs the same OSS spool engine. spool cloud login is what tells it to push to a Spooling Cloud workspace.
Spooling
The open source CLI and local indexer. One Docker command, every session from every AI coding tool you use lands in your own pgvector index. Searchable by you, on your machine, MIT licensed. Free forever.
Install the OSSSpooling Cloud
The multi-tenant cloud the OSS engines push to. Many laptops feed one shared workspace, with cross-developer search, a cross-vendor audit log, SSO, and an MCP endpoint your agents can call.
Join the waitlistMany laptops. One memory.
Each engineer's OSS spool engine indexes their sessions locally. spool push ships them to Spooling Cloud, where they pool into a shared team workspace your admins, agents, and dashboards can read.
Everything your AI sessions
should already do.
Spooling turns the opaque stream of AI coding sessions into a structured, searchable history you actually own.
Semantic search
pgvector + local embeddings mean you find sessions by meaning, not just keywords. Zero API calls, zero data leaves your machine.
Usage stats by tool
Per-session and per-project token counts, tool call histograms, and per-provider rollups, all aggregated automatically.
Git-aware
Sessions are linked to the branch and commit you were on. Trace what you were doing when without guessing.
Cross-tool mesh
Spooling sits between every AI coding assistant and weaves their sessions into one timeline. Switch tools without losing the thread.
Powered by Strands
Built on the Strands agent framework. Traces, spans, and metrics flow through the same observability model AWS ships for production agents.
CLI + GUI
Drive it from your terminal, or browse a local dashboard at localhost. Whichever you're already in.
Four commands.
Zero cloud.
git clone https://github.com/sashimiboi/spooling && cd spooling docker compose up -d # postgres + pgvector :5434
python3 -m venv .venv && source .venv/bin/activate pip install -e . cd ui && npm install && cd ..
spool init # scan for available providers spool sync # embed every session into pgvector
spool ui # API :3002 · MCP :3004 · GUI :3003 spool search "that redis race condition"
The only unified knowledge base
across every AI coding tool.
Each AI coding tool locks its sessions inside its own silo. Spooling is the only tool that reads them all, embeds them into one local vector store, and gives you a single semantic memory across every provider you use.
The glue your AI workflow
is missing.
Spooling lives in the same neighborhood as other AI memory tools, but none of them solve the thing that actually hurts: copy-pasting context every time you switch tools.
| Tool | Primary strategy | Unified knowledge base | Best for |
|---|---|---|---|
| Background sync across IDEs, CLIs, and web chats | Automated | Devs moving between AI coding tools | |
| Isolated project workspaces, chat inside its UI | Only internal chats | Non-devs organizing documents | |
| Manual history hub, searchable archive | Manual sync | Looking up old conversations fast | |
| Local model runner | Session logs only | Running free models on-device | |
| Real-time memory graph over MCP | Live state sharing | Live context hand-off between apps | |
| Cost tracking + API observability | No (traces, not memory) | Ops teams monitoring LLM spend |
Spooling solves AI amnesia by automatically feeding past decisions back into your current session, no matter which tool you opened it in.
Three things a frontier lab
cannot ship.
The AI labs and IDE incumbents will all build per-product usage dashboards. The defensible wedge for Spooling is the work they structurally won't do.
Cross-tool by design
Anthropic, OpenAI, GitHub, and Cursor each ship analytics for their own product. None of them will ever surface cross-vendor comparisons. That's competitive intel against them. An independent third party is the only place that aggregation can credibly live.
Labs won't aggregate across competitors.
Compliance-grade audit trail
Each vendor will give you a usage log for their own product. None will hand you a unified record of every AI-assisted change across all of them. That unified record is what SOC 2, ISO 27001, and FedRAMP reviewers actually ask for.
The audit log only an independent tool can ship.
MCP-portable context
Your AI session history shouldn't be trapped inside whichever assistant you used last quarter. Spooling exposes everything as an MCP endpoint, so the next agent you open can call it as a tool. The data is yours, the assistants are interchangeable.
Bring your AI history with you.
You can't govern
what you can't see.
Procurement is starting to ask hard questions about AI-assisted code. Most engineering orgs can't answer because each vendor only shows its own product. Spooling is the only place a unified answer can live.
Cross-vendor, including the tools you didn't approve
Anthropic Console shows Claude. GitHub Enterprise shows Copilot. Cursor's admin shows Cursor. None of them shows the union, and none of them shows shadow AI your devs picked up last quarter. Spooling reads from disk, so it sees every tool that ever wrote a session file.
Per-developer, per-project, per-prompt
Every prompt and every response is captured and indexed, tagged with the engineer who ran it, the project they were in, and the model that answered. Filter by engineer for performance review evidence. Filter by project for client engagement audits.
Not the vendor auditing itself
A coding tool's own dashboard is its own marketing surface. The numbers it shows you are the numbers it wants you to see. Spooling isn't aligned to any one provider and has no incentive to soften usage attribution.
- SOC 2 CC7.2 (system monitoring): a record of every AI interaction
- ISO 27001 A.12.4 (logging): retention with non-repudiation by user
- HIPAA technical safeguards: per-user audit of accesses to PHI-adjacent code
- FedRAMP AU-2: complete event log across the systems in scope
- Full prompt and full response, not just metadata
- Model, provider, and tool version per turn
- Working directory, git branch, and file context
- Engineer identity (laptop fingerprint, then SSO once Cloud is wired)
An agentic mesh on top
of your unified memory.
Spooling isn't just a passive index. The Spooling Agent sits on top of your unified knowledge base and turns it into an active mesh you can query, reason over, and route work through.
Cross-provider recall
Ask in natural language. The agent retrieves relevant spans from every provider's history and feeds them to the model that's best suited to keep going.
Session handoff
Pick up where any tool left off. The agent reconstructs the working context and hands it cleanly to the next agent, so you never re-explain yourself.
Mesh routing
Route prompts to the right provider based on cost, capability, or past success. Your unified memory becomes the traffic layer for every tool you use.
Your team's AI memory,
unified in one workspace.
Spooling on your laptop is great. Spooling across your team is a different tool. The cloud workspace pools every dev's sessions into one shared mesh.
Shared team memory
Every teammate's sessions land in one pgvector workspace. Search across every prompt, every tool call, every fix.
Onboarding, solved
New hires inherit months of team context the moment they sign in. No knowledge transfer meetings.
Zero ops
We host the Postgres, the pgvector index, the sync layer, and the embedding service. You do not run Docker.
Cross-tool usage visibility
See which AI tools every team member actually uses, rolled up by project, repo, and teammate.
SSO and audit
Enterprise plans get SAML, SCIM provisioning, and exportable audit trails for every session read.
Secrets redacted by default
The CLI strips API keys and tokens before syncing — AWS, GitHub, OpenAI, Anthropic, Stripe, PEM keys, and sensitive env vars all get replaced with [REDACTED] placeholders. Nothing raw leaves your laptop.
Free for solo.
Built for teams.
Spooling Cloud adds shared cross-tool memory, a unified audit trail for every prompt and response, and the SSO controls procurement asks for.
The OSS engine. Runs on your laptop, nothing leaves your machine.
- Every provider: Claude Code, Codex, Cursor, Copilot, Windsurf, Kiro, Antigravity, Gemini, OpenCode
- Local pgvector semantic search
- CLI, GUI, and MCP endpoint
For small agencies and teams trying it out.
- Everything in Open Source
- Cloud workspace with shared pgvector
- Cross-tool, cross-member semantic search
- Cross-vendor audit trail
- 5+ seats, billed annually
For agencies and engineering firms running multiple AI tools.
- Everything in Team
- SSO / SAML, SCIM provisioning
- Audit log export (SOC 2, ISO 27001-ready)
- Per-engineer, per-project AI usage attribution
- 50+ seats, billed annually
For organizations that need SOC 2, BAAs, and an SLA.
- Everything in Business
- SOC 2 reports, BAA available
- Data residency
- Dedicated CSM and SLA
Spooling Cloud is pre-launch. Prices shown are billed annually.
Join the Spooling Cloud waitlist.
We are opening up to a small set of teams first. Leave your details and what hurts most about agentic engineering today.
The things people ask first.
Does my code ever leave my machine?
Not with the OSS version. Everything runs locally. If you opt into Spooling Cloud, only the session text and metadata you choose to sync leaves your laptop, encrypted in transit.
What about secrets? Are my API keys exposed in the cloud?
No. The CLI redacts secrets client-side before anything leaves your machine — API keys (AWS, GitHub, OpenAI, Anthropic, Stripe, etc.), PEM private keys, and sensitive KEY=VALUE lines are replaced with [REDACTED] placeholders. You'd have to pass --no-redact to send raw values.
Do I need an API key to try it?
No. The OSS CLI runs fully offline with local sentence-transformers embeddings. You only need an API key if you want the chat agent to use Anthropic or OpenAI instead of a local Ollama model.
What's the difference between Spooling and Spooling Cloud?
Spooling is the open source engine you run on your laptop. Spooling Cloud is the hosted multi-tenant version where your team's sessions pool into a shared workspace.
Is there a time limit on the open source version?
No. Run it locally as long as you want. No account, no credit card, no usage cap.
Can I self-host Spooling Cloud in our own VPC?
Not in v1. Enterprise self-host is on the roadmap.
Which AI coding tools are supported?
Claude Code, OpenAI Codex CLI, Cursor, GitHub Copilot, Windsurf, Kiro, Google Antigravity, Gemini Code Assist, and OpenCode today. Aider is next.
Start indexing
your AI memory.
Clone the OSS for yourself, or launch a team workspace in Spooling Cloud.