ApolloBot Core
Open source is the engine, not a trial.
Inspect it, run it locally, extend its connectors and keep your evidence. Cloud is the managed operational layer.
Two honest ways to begin.
The TypeScript SDK controls the hosted or local API. The research runtime is Python 3.11+ and installed reproducibly from source.
npm install @apollobot/sdkgit clone https://github.com/frontier-science-ai/apollobot.git
cd apollobot
uv sync --frozen --extra dev
uv run apollo --versionOne project The @apollobot/sdk package and the frontier-science-ai/apollobot repository are the same open-source project. Frontier Science publishes the SDK to npm and hosts the engine source on GitHub.
Apache-2.0
What stays open
- CLI and authenticated worker API
- Research lifecycle and approval checkpoints
- 54 scientific data connectors
- Provenance, artifact and replication formats
- Container sandbox and Cloudflare deployment package
Self-host or managed
Core and Cloud, side by side.
Every capability of the open engine stays yours. Cloud adds the operational layer so a team can run governed research without owning the plumbing.
CoreApache-2.0 · self-operatedCloudManaged operational layer
Source codeApache-2.0 · run, read, and extend the full engineSame engine, operated for you
Compute & storageYour own infrastructure, keys, and regionsIsolated private workspaces and dedicated VMs
Research lifecycleProtocol, approvals, and evidence gates includedIdentical lifecycle, centrally governed
Provider keysBring your own model and connector credentialsManaged runtime secrets, never in payloads
Updates & incidentsYou patch, monitor, and recoverManaged updates, audit logs, and support
Billing & meteringNone — you pay your own providersHard budgets, capacity ledger, guardrails