SessionFS captures every AI coding session and makes it portable across tools. Resume in Codex what you started in Claude Code. Switch machines or come back tomorrow — your agent already knows what happened last session.
Each captured session becomes durable knowledge: decisions, patterns, project context, and wiki pages. Your codebase documents itself over time — agents write back discoveries via MCP.
SessionFS captures AI coding sessions from 10 tools into a portable .sfs record. Switch machines, switch tools, or come back tomorrow without rebuilding context.
Send a session to a teammate, another machine, or an agent persona. The recipient gets the conversation, files, git state, rules, ticket, and provenance instead of a pasted summary.
SessionFS captures the raw evidence first: transcript, tool, workspace state, git snapshot, rules, persona, and ticket provenance. That portable record is what makes resume, handoff, audit, and memory reliable.
Then every captured session is distilled into knowledge entries — architecture decisions, debugging patterns, API contracts, and environment quirks. Your codebase builds its own documentation over time.
AI agents don't just read the knowledge base — they write back. When an agent discovers something new during a session, it adds the finding automatically via MCP. The next agent picks it up.
As your team grows, SessionFS adds coordination, governance, and security — shared knowledge, agent tickets, SSO, DLP, and self-hosted deployment.
Cross-references every AI claim against tool output. Contradictions surface before they reach production.
14 PHI patterns + 20 secret patterns scrub session archives at the server. BLOCK / REDACT / WARN modes.
Every captured session carries instruction provenance, persona, ticket, and the AgentRun that signed off.
Hardened chart (UID 10001, read-only rootfs, RuntimeDefault seccomp). Your data stays in your cluster.
SessionFS starts with capture, then turns that work into reusable context, assigned tasks, and auditable agent runs.
Sessions are captured from 10 tools automatically. The raw work becomes the evidence trail.
The agent receives the right persona, project knowledge, rules, file refs, and acceptance criteria before it starts.
AgentRuns record severity, findings, policy result, and exit code so CI and humans can trust the result.
Every AI coding agent makes claims: "the test passes," "I created the file," "the migration is safe." Most go unverified. LLM Judge cross-references every claim against actual evidence — exit codes, file writes, test output.
Confidence-scored findings surface on PRs and in the dashboard. Your team reviews AI work the same way they review human work.
Verified claims feed the knowledge base — building a trust record your team can reference.
BYOK — your own key, encrypted at rest and never shown again. Works with any OpenAI-compatible endpoint: OpenRouter, LiteLLM, vLLM, Ollama.
Read the docs →An auditable execution record for one persona run — CI-friendly policy enforcement with severity × fail_on → exit_code. Tracking + enforcement, not model auto-spawning. Replay a run with an edited prompt and compare the findings. GitHub Actions + GitLab CI workflows shipped.
Portable AI personas and self-contained tickets with FSM, dependencies, and compiled persona+ticket context. Every captured session is tagged with the active persona and ticket.
Push to the cloud, pull on any machine. Auto-sync available with three modes: off, all, or selective.
Your AI agents read project context, search knowledge, and write discoveries back — all via MCP.
Install an Agent Skill or Plugin once, pinned to an exact version, for every supported coding agent. See exactly what a package can do before you install it, re-approve updates that want more access, get scanned, verified-publisher packages, and see which tools each package is verified to work in. Add a SessionFS catalog with your tool’s own plugin command. Browse the hosted SessionFS Marketplace free.
See what happened in 10 seconds. Files, tests, commands — extracted automatically with no LLM cost. The Changes tab shows a diff for every file the agent touched.
First-class coordination — ticket + persona provenance carries over, sender curates attachments, recipient claims atomically. Revoke / decline / comment threads with audit events. Team handoffs (Team+) let any member claim.
AI session context on every PR and MR, with contradictions surfaced for reviewers. Enterprise adds two-way Jira sync and Confluence publishing.
Deploy to any Kubernetes cluster with Helm. Your data never leaves your infrastructure.
Choose the deployment model that fits your security requirements.
Dedicated infrastructure provisioned for your organization. Isolated compute, storage, and networking. We manage it — you own the data.
Deploy to your own Kubernetes cluster — AWS EKS, GCP GKE, Azure AKS, or on-prem. Nothing leaves your network. Air-gapped deployment available.
Use SessionFS locally at no cost. Add cloud sync, shared knowledge, tickets, AgentRuns, service keys, and enterprise deployment as your team adopts AI agents.
Transparent pricing for solo developers, teams, and self-hosted enterprise deployments.