Skip to content

early access

Project cycles for teams where development never stops

Build and maintain with confidence by orchestrating task agents through repeatable cycles. vem revalidates completed work, reduces accidental breaking changes, and helps ensure features keep working as intended.

no `.vem` in gitpending to verifiedMCP + CLI parityWorkspace terminalSandbox task runsContinuous project cycles
Cycle ModelVerified Cloud Memory
Orchestrate task agents
scoped implementation cycle
Push cycle snapshot
snapshot saved as pending
Revalidate on git events
snapshot promoted to verified
Cycle SnapshotCloud DB
{
  "cycle": "revalidation-pass",
  "snapshot_hash": "f74b...",
  "git_hash": "ad71...",
  "status": "verified"
}

Cloud DB source of truth

Shared state lives in the cloud, not in git-tracked `.vem` files.

Git-backed verification

Each snapshot carries `git_hash` + `snapshot_hash` for auditability.

Fast team retrieval

Search and analytics run against one canonical memory stream.

Secure by default

Tenant isolation and traceable timeline for memory updates.

Shared memory

Shared memory for every AI agent in your repo

Keep `.vem/` as a local workspace cache and sync canonical memory to the cloud. Tasks, decisions, and snapshots become searchable, auditable, and available to every agent and teammate.

Web execution

Operate repos from the web without losing execution boundaries.

vem splits online execution into two clear paths: paired local workspace commands and isolated sandbox task agents.

Workspace terminal

Web commands executed by a paired vem runner

  • Start a runner on any machine: vem runner --agent copilot
  • The runner reports its online status, current branch, and worktree state back to the web UI in real time.
  • Submit normal shell commands from the browser and see streamed output — ideal for git status, build checks, or diagnostics.

Cloud task execution

Task agents run in isolated sandbox workspaces

Ultra
  • Launch an agent directly from a task page — no local runner required.
  • The sandbox clones your repo from GitHub into an isolated environment before execution.
  • Successful runs push a task branch and open a pull request automatically.

Project Cycles

Development and maintenance never stop. Revalidation should not stop either.

Project cycles keep delivery honest: each pass rechecks prior work, orchestrates task agents around scoped changes, and reduces the risk of accidental regressions before merge.

Step 1

Revalidate completed work

Run the same standards against existing features so drift and regressions are surfaced early.

Step 2

Orchestrate task agents

Coordinate focused agents per task so implementation, checks, and handoffs stay explicit and traceable.

Step 3

Confirm intended behavior

Use cycle feedback to verify features still behave as intended while reducing risky, accidental breakage.

How It Works

Three steps from context to verified memory.

The workflow is built for high-velocity teams: capture locally, sync immediately, verify automatically with git events.

Step 1

Capture locally

Agents and developers update `.vem` as local workspace memory.

Step 2

Push to cloud

`vem push` writes pending snapshots to the shared memory service.

Step 3

Verify on git push

Webhooks match provenance metadata and finalize snapshots as verified.

Storage Model

Keep local memory lightweight. Keep shared memory canonical.

Local files help the current workspace. Cloud records keep the durable team timeline. This separation removes git noise and keeps collaboration consistent across agents.

Default behavior

`.vem/` is gitignored and used as a local cache for active work.

Cloud state powers search, analytics, and cross-agent handoffs.

Git metadata verifies integrity without committing workspace memory.

Local workspace cache (`.vem/`)

Current context and active task objects

Fast local edits by CLI and MCP tools

Not committed to git

Canonical cloud memory

Pending and verified snapshot lifecycle

Searchable tasks, decisions, and context history

Shared state across humans and agents

Why Teams Choose vem

Memory you can trust under real delivery pressure.

vem aligns multi-agent workflows with the same source of truth, so handoffs are faster and regressions are easier to prevent.

Included in the platform

MCP tools + CLI parity

Task + decision tracking

Snapshot verification lifecycle

Cloud search + analytics

Workspace terminal + runner

Sandbox task runs + PR creation

Agent + human parity

MCP tools and CLI operate on the same shared memory state.

Clear orchestration

Track intent, evidence, and handoffs without prompt archaeology.

Git-verified timeline

Every snapshot is tied to deterministic provenance metadata.

Fast retrieval

Find relevant decisions and tasks in seconds across project history.

Workspace terminal

Run project commands from the web through a paired local runner.

Task to PR flow

Trigger isolated task agents online and turn successful runs into branches and pull requests.

Cloud indexing

Query and analyze canonical memory across repos and teams.

Secure by default

Tenant isolation and auditable trails for critical workflows.

Start Building

Move from fragile prompts to durable memory.

Give your team one place for verifiable context, shared history, and fast retrieval across every coding session.