One engineering brain for all your AI coding agents
Qorvyn gives Claude Code, Codex and any coding agent the same cited understanding of your project. Describe a goal; specialist agents plan it, build it, review it and test it. You approve, then the change is delivered. Agents run on your own computers, next to your code.
$ npx qorvyn initEverything your agents need to work like a team
Bring the agents you already use. Qorvyn adds the shared memory, the process and the safety around them.
Cited project context
An incremental index of your repository. Every agent gets the same answers with path:lines@commit citations, through the CLI, an SDK or MCP.
Describe a goal
A planner turns your goal into a workflow of steps, reusing workflows you already saved. You watch it think, then approve the plan.
Any agent, any model
Claude Code and Codex built in, plus any CLI agent (Cursor, Gemini CLI, Aider, OpenCode) or chat API. Pick the model per agent. Keys are write-only.
Guardrails, enforced
Denied capabilities, protected paths, allowed providers, time limits and approvals for every risky action. Enforced by Qorvyn, not by asking the model nicely.
Safe delivery
Agents commit to their own branches, never your working folder. Get the changes locally, merge, or open and merge pull requests on GitHub, GitLab, Azure DevOps or Bitbucket.
Products across repositories
Group a backend and its app into one product. Describe a goal and the planner spreads it across the repositories; each step runs in its own repository, under that project's guardrails, and each repository delivers its own changes.
Team sharing, live
Share agents, workflows and knowledge through a server you host. Signed changes, live updates, conflict detection and an offline queue.
Knowledge that learns
Add documents (PDF, Markdown, text) and search them with citations. Runs propose lessons, decisions and discoveries; only what you approve reaches agents, and knowledge resting on files goes stale when those files change.
Products from idea to production
A product can start before any repository: an objective and a versioned spec (requirements, acceptance criteria, components) that you approve. See what a change touches, and how the code, CI and deployments compare with the spec.
Evidence, not claims
Qorvyn runs your checks itself, on the exact commit, in a clean worktree or a container with no network. Releases are ready only on evidence; secret, dependency and code scanners run as checks.
Initiatives
Big work as one initiative: a workflow can run other workflows, never with more approval than its step has. See progress, blockers and one timeline; steps wait for a provider or a budget instead of failing, and failed work can be rolled back.
Cost before you run
Forecast the AI cost of a workflow before it starts, from your measured runs. Budgets for runs and initiatives are enforced before every step, and each step's model choice is recorded.
Pipelines and Fix with Qorvyn
Your GitHub Actions runs with the failing lines of each failed step. Re-run, cancel or start a pipeline, and one click has an agent fix the failure in a pull request, never by loosening tests.
Environments and deployments
Plan Azure environments with prices, approve them within a budget and apply them once through your own sign-in. Canary deployments roll back automatically when health checks fail; backups and an incident workflow included.
Autonomy you choose
Supervised, Assisted or Autonomous. Production deployments, secrets, pushes and cloud changes always wait for a person. The Engineering page shows initiatives, verified success, cost and evidence.
From goal to reviewed change in four steps
Qorvyn sets itself up from what your repository already has: CLAUDE.md, AGENTS.md, your existing agents and skills.
Set up
npx qorvyn activateonce for your free account, thennpx qorvyn initindexes the repo and creates read-only agents, workflows and knowledge for you to review.Describe a goal
The planner reads the code, designs the steps and shows the plan before anything runs.
Agents work
Steps run in parallel; agents that change files work in their own worktrees. You see live what each agent reads and changes.
You decide
Review the diff, approve, and deliver: into your branch, or as a pull request.
Your code never leaves your machine
The index, runs and notes live in ~/.qorvyn, outside your repository. A team server receives only what someone chooses to share, signed by their device, and the results of work you choose to run through it.
- Installing adds no files to your repository: only the dependency entry in
package.jsonandpackage-lock.json. - Sensitive files (.env, keys, credentials) are never read or indexed.
- API keys are write-only: in the macOS Keychain, or an encrypted file on other systems.
- Anonymous weekly usage counts, opt-out and documented field by field:
qorvyn telemetry off.
# set up the project (local database, index, agents) npx qorvyn init # ask any question about the code, with citations npx qorvyn context search "authentication middleware" # describe a goal: plan, build, review, test npx qorvyn do "rate limit the login endpoint, with tests" # open the dashboard npx qorvyn dashboard
Core for everyone. Enterprise for organizations.
The same product and the same Q. Enterprise is a build of the same team server, not a fork, so developers keep using Core.
CoreQorvyn Core
- Cited context, CLI, SDK and MCP server
- Agents, workflows, Describe a goal, dashboard
- Any provider and model; write-only keys
- Guardrails and safe delivery to every git host
- Self-hosted team sharing with roles and live sync
- Runners and a control plane: plans, approvals, verified results, budgets
- Knowledge from documents and runs, approved before agents use it
- Product specs, evidence-based releases, cost forecasts, Azure environments
- Tamper-evident audit log
EnterpriseQorvyn Enterprise
- Everything in Core
- Organization policies enforced on every device, including a ceiling on agent autonomy
- SSO group roles and SCIM provisioning (Entra ID, Okta…)
- Audit export, retention and signed SIEM webhooks
- Runner pools in your own Azure, secrets from your vault, spending rules and assurance levels
- Signed audit checkpoints, proofs and legal holds
- Central seats, device revocation and live entitlement enforcement
| Capability | Core | Enterprise |
|---|---|---|
| On every developer's machine | ||
| Index, cited search, CLI, SDK, MCP | ✓ | ✓ |
| Agents, workflows, planner, live progress | ✓ | ✓ |
| Providers and models, write-only keys | ✓ | ✓ |
| Guardrails and delivery (GitHub, GitLab, Azure DevOps, Bitbucket) | ✓ | ✓ |
| Pipelines (GitHub Actions) and Fix with Qorvyn | ✓ | ✓ |
| Knowledge: documents, lessons, decisions, validated and kept fresh | ✓ | ✓ |
| Products and specs, change impact, initiatives with child workflows | ✓ | ✓ |
| Checks run by Qorvyn as evidence (worktree or container sandbox), release readiness, living documentation | ✓ | ✓ |
| Cost forecasts, budgets before every step, recorded model choices | ✓ | ✓ |
| Environments on Azure: priced plans, canary deployments with rollback, backups | ✓ | ✓ |
| Backlog, defect register and manual tests; client progress reports; plan roles that approve | ✓ | ✓ |
| Autonomy levels (supervised, assisted, autonomous) and the Engineering page | ✓ | ✓ |
| On the team server | ||
| Workspaces, roles, invitations, live sharing | ✓ | ✓ |
| Sign-in with Microsoft Entra ID, Google, Okta (OIDC) or email codes | ✓ | ✓ |
Control plane and runners: plans with approvals, leases, verified results, run budgets (API and qorvyn runner; dashboard pages coming) | ✓ | ✓ |
| Tamper-evident audit log (hash chain, verify on demand) | ✓ | ✓ |
| Products in the workspace, specs encrypted with the workspace's own key | ✓ | ✓ |
| Organization policies on every device, including the highest autonomy and write access for new agents | — | ✓ |
| SAML 2.0 sign-in, SSO group roles, SCIM provisioning | — | ✓ |
| Audit export, retention, SIEM webhooks | — | ✓ |
| Seats, deactivation and live entitlement enforcement | — | ✓ |
| Dedicated runner pools, minimum trust, workspace quotas | — | ✓ |
| Required budgets, four-eyes spending approval, monthly budgets, cost reports | — | ✓ |
| Assurance levels (enhanced, high) | — | ✓ |
| Signed audit checkpoints in immutable storage, proofs, legal holds | — | ✓ |
| Secrets from your vault for runners (Azure Key Vault references, audited) | — | ✓ |
| Runner pools in your own Azure (approved, priced plans; one token per pool) | — | ✓ |
| Managed runners and AI in Qorvyn's cloud, hosted cloud | — | Planned |
Bring cited project context into any stack
The TypeScript SDK ships with Qorvyn today; more languages are on the way. Your code and index remain on this machine.
Up and running in a minute
Needs Node.js 22.13 or newer and a free account (one command). No server or database to set up. Qorvyn is on npm; the source is not public.
- Works with Claude Code and Codex out of the box
- Monorepos understood: apps and packages detected
- Add a team server later; nothing to migrate
# in a JavaScript/TypeScript project (or: npm install -g qorvyn, for any repository) npm install -D qorvyn npx qorvyn activate # once: your free account npx qorvyn init npx qorvyn dashboard
Questions, answered
Does Qorvyn replace Claude Code or Codex?
No. Qorvyn works with them. It gives every agent the same project understanding, runs them as specialists in workflows with approvals, and keeps them inside your guardrails.
Does my code leave my machine?
Qorvyn itself never sends your code anywhere. The agents you run talk to their own providers (for example Anthropic or OpenAI) under those providers' terms, with the context Qorvyn chooses: cited snippets, never sensitive files. A team server only receives what you share, and the results of work you run through it.
Will it add files to my repository?
No. Only the dependency entry in package.json and package-lock.json. Settings and data live in ~/.qorvyn. Agents work in their own worktrees and commit to their own branches.
Is Qorvyn free?
For individuals, yes: a free account (one command) with up to three personal projects, and it does not expire. Teams and organizations that need more, and Enterprise (per seat, with governance, audit evidence and SSO), are on paid plans: contact us.
What is the difference between Core and Enterprise?
Core is the product for individuals and teams, with self-hosted team sharing. Enterprise is a commercial build of the team server that adds organization policies, governance of runners and spending, assurance levels, signed audit checkpoints, SSO groups, SCIM, audit export and seats. Developers keep using Core.
Does it work offline?
Your code, index and notes are local, but Qorvyn checks your account online before protected commands: a short signed request with your device, project slot and the command name, never code, prompts or file names. So it needs an internet connection, and there is no offline grace period.
Can agents run on other computers?
Yes. Add a runner from the dashboard's Runners page (a second computer, a server, a VM in Azure or AWS, a CI machine): it gets a one-time token, and the team server hands it approved tasks. Send any of your workflows to your runners from the same page. It runs them with its own project agents, guardrails and checks. A task never carries a command: the checks a runner runs are configured by its owner.
Do I need a server?
Only to share with teammates or run work on other computers. Try it on your own machine with npx qorvyn server start (it runs its own database), or run it in your cloud (for example Azure, with PostgreSQL) for the team.
Is there telemetry?
Yes, anonymous and opt-out: weekly counts of which commands and pages are used and how runs end, with version and OS. Never code, file or repository names, prompts or outputs, and no IP addresses are stored. See every field, or turn it off with qorvyn telemetry off or DO_NOT_TRACK=1. Details: Data governance.
Give your agents one brain
Start with Core today. Talk to us about Enterprise for your organization.