Audit-readyAir-gappedMulti-user Hub

ResearchForge
for research teams that need control

When experiments directly affect product decisions, the critical question is no longer whether the result was good — it is whether the team can prove how it was produced, who reviewed it, and what changed in the process. ResearchForge Enterprise adds that layer.

Book a 30-min demoRead enterprise docs →

You know this problem.

Today — without ResearchForge
✗Someone ran "the best model" last quarter but nobody remembers the config
✗Three engineers ran conflicting experiments on the same branch
✗The baseline shifted between runs — "improvement" is meaningless
✗An intern touched a protected file and nobody noticed until prod
✗Compliance asked for a full experiment audit — you have a Slack thread
✗Good hypotheses die because testing takes 3 days to set up
With ResearchForge Enterprise
Full experiment lineage in the Hub — every result, every config, forever
Git worktree isolation — parallel experiments, checkout always clean
Frozen baseline at commit time — reproducible from day one
Cryptographic protected-path contract — violations kill the run immediately
Immutable SOC2-ready audit log — exportable JSON, user-attributed
Hypothesis → experiment in one command — no setup overhead

The open-source CLI, plus the structure your team needs to scale.

Enterprise adds the shared controls, governance, and coordination layer that makes research reproducible across a real team without slowing down individual contributors.

Enterprise
🔒

Air-gapped VPC deployment

Hub server runs inside your network. No data leaves. arXiv search optional and disableable.

Enterprise
👥

Multi-user Hub dashboard

All team experiments in one place. Live run status, lineage DAG, approval queue.

Enterprise
⚡

CI/CD plugin

GitHub Actions & GitLab CI. Experiments triggered on every PR, results posted as comments.

Enterprise
📊

Private tracker integrations

Mirror validated results into your existing reporting and observability workflows without breaking local-first research habits.

Enterprise
🤖

Private model routing

Use your own model endpoints for hypothesis generation and synthesis with an OpenAI-compatible interface.

Enterprise
📋

SOC2-ready audit trail

Immutable log of every action, every user, every result. JSON export on demand.

Enterprise
🛡️

Protected path enforcement

Cryptographic contract. Any experiment that touches protected files is killed and logged.

Enterprise
☁️

Cloud execution adapters

AWS Batch & GCP Cloud Run workers. GPU-tagged routing. Scale beyond local machines.

Two lines of config. Works with everything you already use.

Use the same CLI you already trust, then point it at a shared Hub when your team needs visibility, approvals, and consistent experiment tracking.

~/.researchforgerc
# Before (open-source, local only)
# — nothing needed —


# After (enterprise, syncs to Hub)
RESEARCHFORGE_HUB_URL=https://hub.yourco.com
RESEARCHFORGE_API_KEY=rf_live_xxxxxxxxxxxx

Same CLI commands. All results now sync to the shared Hub.

🏗️
Hub stays in your VPC
One docker run command. Your network, your data, your control.
🔑
API key per team member
Revocable, scoped tokens. Fine-grained permissions per workload.
🔗
Workload tagging
Tag experiments by project. Hub groups and compares across team runs.
📡
CI/CD in 5 lines
GitHub Actions workflow included. Experiments on every PR, auto-commented results.

Plans

FeatureOpen SourceTeamEnterprise
PriceFree foreverContact usCustom contract
ExperimentsUnlimitedUnlimitedUnlimited
Users / hub1 (local)Multi-user hubOrg-wide dashboard
IDE supportClaude Code + Cursor+ CI/CD plugin+ Custom LLM
ExecutionLocal venv / DockerLocal + Cloud runnersAir-gapped / VPC
Auth—SSO / SAML readyOkta, Azure AD
Audit trailGit logFull API auditSOC2 / ISO27001
Integrations—Slack / TeamsMLflow, W&B, custom
SupportCommunityEmail SLADedicated + SLA

Frequently asked

Is the individual / open-source version really free forever?▾

Yes. The full CLI — research search, baseline, run, validate, ship — is Apache 2.0 and free forever for individuals and teams up to 2 people. No nags, no feature flags, no expiry.

What does "air-gapped" actually mean here?▾

The Hub server runs as a Docker container inside your VPC. No results, code, or metadata leave your network. The only optional external call is arXiv for paper retrieval — which can be disabled and replaced with a pre-populated local cache.

How does the Hub fit into a team's existing stack?▾

The Hub is designed as a control plane around your existing workflow. It works with your private model endpoints, CI pipelines, team reporting, and local execution setup without forcing a vendor lock-in.

Can we use our own LLM instead of Claude or Cursor?▾

Enterprise tier supports custom LLM endpoints via the OpenAI-compatible API. Point researchforge at your private fine-tuned model and it drives the hypothesis generation and literature synthesis loop.

How is the protected-path contract enforced?▾

Before any experiment runs, ResearchForge hashes all protected paths and commits the hash to the experiment contract. After the run, it re-hashes and compares. Any divergence — even a single byte — kills the experiment and logs a violation.

What does the audit trail cover?▾

Every command run, every experiment start/stop, every metric emitted, every approval or rejection decision, and every ship action is logged with a timestamp and user identity. The log is exported as structured JSON on demand.

Ready to bring reproducibility to your team?

Book a 30-minute demo and we'll walk you through the Hub setup, CI/CD integration, and air-gap options for your infrastructure.

Book a 30-min demosupport@researchforgehq.com
Open-source on GitHubRead the docs →