Cogniplane Core is open source.
Cogniplane Cloud is the managed version.
Run the AGPL-3.0 release on your own infrastructure, or let us run it for you. Predictable per-capacity pricing, SSO, audit retention, dedicated support. Same backend either way. Pick whichever operating model fits your team.
You'll need Docker, Node 24, one model-provider API key, and an
E2B key for the code-execution sandbox
(make e2b-build builds the template in your own account). The backend
refuses to start without them, on purpose. Full steps are in the
README.
For enterprise teams who want the platform, not the operations.
Self-hosting Cogniplane Core means you run the sandbox fleet, the database, the secret rotation, the upgrade path, and the on-call rotation. Cogniplane Cloud is the same backend with all of that included, so your team can focus on what the agents do instead of keeping them running.
We run the sandbox fleet
E2B sandboxes scale automatically with load, with the isolation, networking, and capacity you'd otherwise build yourself.
We handle updates
Security patches, runtime updates, and feature releases roll out continuously. No upgrade windows for your platform team.
Enterprise controls included
SSO via WorkOS, audit log retention, dedicated CSM, configurable approval policies, and an SLA on the Enterprise tier.
Same platform, different operating model.
Cogniplane Core is licensed under the AGPL-3.0. If you're comfortable with the AGPL's network-service obligations and want to run it yourself, the OSS release is feature-complete for the core platform. The governance layer is not held back: self-hosters get the same Policy Center rule engine, the same approval gates, the same multi-tenant isolation, and the same audit trail we run in Cloud. What Cloud adds is the managed integrations, the enterprise services, and not having to operate any of it.
| Feature | Self-host AGPL-3.0 · Cogniplane Core $0 | Cloud Professional Most teams $1,999 / mo | Cloud Enterprise VPC · self-hosted options Custom |
|---|---|---|---|
| License | AGPL-3.0 | Commercial | Commercial · custom DPAs |
| Cost | Free (you run the infrastructure) | $1,999/month | Custom |
| Concurrent sessions | Bring your own | 20 included | 50–500+ |
| Infrastructure operations | Your team | Included | Included |
| Updates & security patches | Manual git pull + redeploy |
Continuous | Continuous |
| Model providersAnthropic, OpenAI, Google, OpenRouter, Z.AI | All five · your own API keys | All five · your own API keys | All five · your own API keys |
| SSO (WorkOS) | Self-configure | Included | Included |
| GitHub integration | ✓Included | ✓Included | ✓Included |
| Microsoft / SharePoint integration | —cloud-only connector | ✓Included | ✓Included |
| Scheduled jobs | ✓Included | ✓Included | ✓Included |
| Approval workflowsNative runtime approvals plus the full Policy Center rule engine | Full (granular policies) | Full (granular policies) | Full (granular policies) |
| Private PII detectionPII never leaves the platform — no third-party model, no query logs | Self-managed (BYO model — your provider's logging applies) | Add-on · private local model | Included · private local model |
| Audit log retention | Self-managed | Standard | Custom |
| Support | Community (GitHub issues) | Email + chat, 24h SLA | Dedicated CSM, 4h SLA |
| SLA | — | — | 99.9% uptime |
| Compliance | DIY | — | SOC 2 PLANNED · custom DPAs |
| Deployment | Anywhere | Cloud (multi-tenant) | Cloud, VPC, or self-hosted |
Private PII detection on Cogniplane Cloud means exactly that: PII is identified by a Cogniplane-operated local model running inside our infrastructure. The detection request never reaches a third-party inference API, and the model does not retain query logs. PII never leaves the platform's data perimeter. The OSS release ships the same code path, but routes through whatever model provider you configure, so your provider's data-handling and logging policies apply. Microsoft connectors are exclusive to Cloud.
FAQ.
What is a concurrent session?
A concurrent session is one agent session running at the same time. If you have 20 concurrent sessions, 20 team members can run agents simultaneously. Others queue until a slot opens. Concurrency is always-on capacity, not a consumable. Think of it like lanes on a highway, not gallons of gas.
Why per-capacity instead of per-seat?
Most enterprise teams have far more potential users than simultaneous active users. A 500-person company might need only 20 concurrent sessions, but paying per seat would 25× the cost for no actual usage difference. Per-capacity is honest pricing tied to our real cost driver (the compute behind active sessions), not to your headcount.
What's the difference between Cogniplane Core and Cogniplane Cloud?
Cogniplane Core is the open-source platform under AGPL-3.0. You can run it on your own infrastructure, modify it, and deploy it however you want, subject to the AGPL's network-service source-disclosure obligations. Cogniplane Cloud is the managed enterprise version: same backend, plus Microsoft connectors, managed PII detection, an SLA, dedicated support, and we run the infrastructure for you. Pick the operating model that fits.
Do you offer a free tier or trial of Cloud?
No free tier. If you want free, the OSS release is feature-complete for the core platform. Download it and self-host. Cloud is for organizations that want the managed experience with enterprise SLAs and integrations.
Can I scale capacity up temporarily?
Yes. Add concurrent sessions on-demand from the dashboard. Billed pro-rata for the remainder of the billing cycle. Volume discounts apply automatically beyond Professional's included 20.
What does the AGPL-3.0 require if I self-host?
The AGPL-3.0 requires that if you offer a modified version of Cogniplane Core as a network service to others, you make your modifications available under the same license. If your use is internal, or you don't modify the source, this obligation is trivially satisfied. If you can't or don't want to comply with the AGPL (for example, building a proprietary closed-source product on top of Cogniplane), Cogniplane Cloud or a commercial license is the alternative. Commercial licensing details.
What does "private PII detection" actually mean?
PII detection on Cogniplane Cloud runs against a Cogniplane-operated local language model. Not OpenAI, not Anthropic, not OpenRouter, not any third-party inference API. The detection request stays entirely inside Cogniplane infrastructure, and the model does not retain query logs. So PII flagged by the detector never leaves the platform's data perimeter and is never recorded by anyone downstream. This matters in industries like healthcare, finance, legal, and defense, where data-handling regulations or contractual commitments forbid sending personal information to external inference APIs. That kind of enterprise often has to disable PII detection entirely on other platforms.
Where does my data live in Cloud?
Cogniplane Cloud runs in us-east-2 (AWS) by default, with Postgres on Neon, Redis on Upstash, and S3 for artifacts, all in the same region. Enterprise customers can request alternative regions or VPC deployments; talk to us about data-residency requirements.
How does support work?
Self-host: community support via GitHub Issues. Cloud Professional: email + chat with a 24-hour SLA. Cloud Enterprise: dedicated CSM with a 4-hour response SLA. Security issues across all tiers go to security@cogniplane.ai per the project's security policy.