AI workspace and runtime

hecate

Run Hecate on your machine as the personal control plane for AI work: projects, chats, model providers, memory, approvals, diffs, traces, and supervised external agents in one UI.

v0.5.0-alpha.5· release notes·all releases· GitHub

public alpha

Install the local console.

Start with the desktop app, Docker image, or source checkout. Hecate runs locally by default and keeps the runtime boundary visible.

project workflow

One workspace for personal and project agents.

Start with a project, attach workspace roots when files matter, and keep chats, tasks, external-agent sessions, context, memory, and evidence connected to the same work.

Chats and routingTalk to models through OpenAI-compatible and Anthropic-shaped APIs with health, failover, and usage visibility.
Projects and memoryKeep work areas, workspace roots, assignments, context snapshots, evidence, and approved memory together.
Supervised agentsRun Codex, Claude Code, Cursor, and Grok Build from Hecate while their CLIs keep their own accounts and runtime behavior.
Observable workKeep approvals, traces, route reports, artifacts, diffs, logs, usage, and final output close to the decision that produced them.
Hecate chat screen with an approval required before a file write

operator control

Agents wait before crossing the line.

Hecate keeps file writes, tool actions, approvals, artifacts, and run context visible before supervised work changes the workspace.

observability

Route reports and traces are first-class.

Hecate is OpenTelemetry-first: requests surface trace IDs, route decisions, timing, token usage, and provider-reported cost where available.

Hecate observability screen with recent requests and spans

remote option

Use Hecate away from your desk.

hecate.sh is the open-source Hecate runtime and app. Hecate Cloud is the separate hosted deployment and access layer for opening that same Hecate UI from a browser.

Hecate Cloud