Plan, code, build, deploy, and monitor — unified in one workspace. When everything shares one data model, AI finally has the context to do real work.
curl -fsSL https://proxifai.com/install.sh | sh
Install the pfai CLI on macOS & Linux · self-update any time with pfai update
The average engineering team uses 10+ disconnected tools. Context is scattered, workflows break at the seams, and AI tools are useless — because they can only see one silo at a time.
Four bands, fifteen sections, one data model — this grid is the app's actual sidebar. When planning, code, deploys, and operations share one workspace, AI stops being a bolted-on feature.
A scope-aware cockpit: org health, a ranked “needs you now” worklist, and every service at a glance.
Reviews, deploy approvals, incidents, and mentions in one triage queue — approve a deploy without leaving.
Everything assigned to you, with the same list and board tooling as Plan. Keyboard-first triage.
Issues, sprints, roadmap, and real-time collaborative docs — connected to the branches and PRs that ship them.
A native git forge: pull requests with inline review, CI checks in the merge box, releases, packages, and an in-browser IDE.
GitOps deployments, environments, config & secrets, and per-PR preview environments that clean up after themselves.
Visual DAG automation: triggers, approval gates, agents in isolated VMs, live execution timelines, a dead-letter queue.
Managed vClusters, Postgres with a built-in SQL IDE, Kafka, and S3-compatible object storage.
A zero-config operations dashboard, custom dashboards, log exploration, and a compliance-grade audit trail.
Incidents with escalation and one-click acknowledge. Rules over metrics, logs, webhooks, and synthetics.
Spend rollups and allocation, budgets with hard stops, anomaly detection — and a “saved by proxifai” ledger.
Real DORA metrics and SLOs with error budgets. Honest bands, not vanity grades.
An assistant that operates the platform: reads issues, searches code, runs commands, opens PRs — streamed live.
One RAG index over docs, code, and chat. Instant, semantic, or hybrid search — the assistant retrieves from it too.
An OpenAI- and Anthropic-compatible LLM gateway: usage analytics, budgets, rate limits, and BYO provider keys.
2 PRs are blocking API v2:
#139 #142 Both are in Sprint 15. I can nudge the reviewers, or take the security review myself in Develop mode.
Assign a task to an AI agent. It reads the issue, understands the codebase, writes the implementation inside an isolated VM, runs tests, and opens a PR.
AI assistants can only be as good as the context they have access to. When your data is spread across ten products, no amount of plugins can stitch it together.
Every action — whether taken by a developer or an AI agent — flows through the same RBAC, audit logs, and compliance controls.
One account for everything — projects, repos, CI, AI, agents. SSO for the entire platform.
Roles, service accounts, guardrail ceilings, and just-in-time elevations — the same bindings govern developers and agents.
Every commit, deploy, and AI action in one decision log — with SOC-2-ready compliance exports per service.
Replace five invoices from five vendors with one plan. Costs go down, not up.
pfai is a single command-line tool that talks to the entire platform. Plan issues, push code, trigger workflows, query AI, and launch agents — all from your terminal.
One login, one API, one token. The same interface your AI agents use to automate everything.
$ curl -fsSL https://proxifai.com/install.sh | sh
macOS & Linux, no dependencies. Already installed? Run pfai update to upgrade to the latest release.
A beautiful UI means nothing if it goes down on deploy day. We build for the worst case first — then make it fast, then make it pretty.
Replicated control planes, distributed storage, and automated failover. If a node goes down, the cluster reschedules workloads without human intervention.
Database replicas, self-healing nodes, and rolling deployments. When hardware fails — and it will — the platform recovers without you noticing.
From a 3-person startup to a 500-engineer org. Kubernetes-native architecture means adding capacity is infrastructure, not a rewrite. No performance cliffs.
Public status page with real uptime data. When something breaks, we post root cause analyses, not vague updates. We'd rather tell you the truth fast than look good slowly.
Continuous database backups with point-in-time recovery. Encrypted at rest and in transit. Your code and data survive anything short of the building burning down — and even then, the other region has it.
Real engineers, not chatbots. When your deploy is stuck at 2am, you'll talk to someone who can read the logs and fix it — not someone reading from a script.
Flashy demos get attention. Uptime gets trust. We care about both, but we know which one matters at 3am on a Friday.
Cloud and self-hosted run the same codebase — vetted, hardened, and operated by us, or deployed on your own infrastructure under an enterprise license.
Deploy on your own infrastructure under an enterprise license. Full control, no vendor lock-in, no data leaves your network.
The same codebase — vetted, hardened, and operated by the team that wrote it. Every release is tested and security-patched before it reaches your data.
One codebase, two ways to run it. Cloud or self-hosted — vetted, patched, and production-ready.
Replace the tool sprawl with a single workspace where planning, code, CI, and AI work together from day one.