Connect all your providers
Anthropic, OpenAI, Fireworks, all sit behind one shared, authenticated AI endpoint.
The open intelligence platform you own.
Model routing, inference policy, and zero-instrumentation telemetry with tapes, deployed in your own cloud
The platform for AI enablement.
Give your team an easier way to use AI while your organization decides what happens behind the scenes.
Capture high fidelity traces
Sessions
| Name / Model | |
|---|---|
Explore semantic routing GPT-6.1 Sol | Alex Chen |
Review authentication flow Claude Opus 5.5 | Jordan Lee |
| Update the product siteGPT-5.6 | Sam Rivera |
Know who ran each session
Paper proxies for attribution and attaches the verified user ID to the session, so each session has a named owner.
Compatible with OpenTelemetry
Manage access to models
Select which models your organization uses. Give an allowlist of exact model names, or let all models through.
Optimize how your teams use AI.
Every session your team runs is saved as structured history, not terminal output. It includes each prompt, turn and tool call, plus the model, tokens, cost, duration and completion status.
Point Paper at sessions that worked, and it generates the skill from what actually happened: the real commands, errors and fixes. One engineers debugging work then becomes a skill the whole team can reuse.
Paper runs real prompts from your recorded sessions twice, once with the skill and once without it, and grades both runs against the same checks. You see how much the skill actually changes results, whether it helps or hurts, and when its quality drops over time.
Insights answers one question: where are the time and money going? Track spend, agent time, tool calls and completion rate across your team. Paper flags sessions that cost far more than usual or did not finish
Get a short summary of what your teams agents worked on, without opening each session. See which work repeats across the team, because repeated work shows you a skill your team does not have yet.
Open source
Build on open foundations.
tapes captures the data layer. Cassettes extend it with independent APIs. stereOS provides a hardened operating system purpose-built for the agents generating that data.
01
tapes
Full-fidelity telemetry for AI agents. Capture, inspect, search, and replay the requests and responses behind every session.
View on GitHub
02
ExtensionsPublic cassettes
Skills cassette
Generate, store, version, and serve reusable skills extracted from tapes trace data.
View on GitHub →Export cassetteExport one session or a time range as JSONL with trace- or span-level detail.
View on GitHub →Search cassetteAdd semantic span search, an MCP search tool, and background embeddings to tapes.
View on GitHub →Memory cassetteStore reviewed memories from captured sessions, then recall accepted entries over HTTP or MCP.
View on GitHub →03
stereOS
A hardened, minimal Linux operating system built on Nix and purpose-built for AI agents, packaged as reproducible machine images called mixtapes.
View on GitHub →Latest
From the blog
BekahHWThoughts · October 7, 2026
Using the right AI should be the easy part
Read post
Start with paper
Turn every session into knowledge at team scale.