// comparison

ChronoMCP vs. saga orchestrators

Short version: they solve overlapping but different problems, and they are usually complementary. Orchestrators like Temporal, Restate and LangGraph implement compensation inside your own workflow code. ChronoMCP is a transparent guard at the protocol level (MCP) — it works with any agent and any tool, adds human approval and an honest view of what cannot be undone, without you rewriting anything.

The one-paragraph difference

A saga orchestrator is a framework you build your workflow on top of: you write the steps and their compensations in code, and the engine runs them durably. ChronoMCP is a proxy that sits between an AI agent and its tools (over the Model Context Protocol) and governs each tool call — previewing impact, gating risky mutations on a human, rolling back declared-reversible steps on failure, and telling the truth about the irreversible ones. You don't build your agent "on" ChronoMCP; you put it in front of whatever agent you already have.

  ChronoMCP Orchestrators
(Temporal, Restate)
Agent frameworks
(LangGraph)
DIY / native MCP
annotations
Where compensation livesProtocol-level proxy (transparent)Inside your workflow codeInside your agent graphAd-hoc in your code
Works with any MCP server unchangedYesN/A (not MCP-specific)Depends on your graphManual, per tool
Human approval gate + impact diffBuilt-in (terminal, policy, Slack quorum)No (you build it)NoNo
Honesty about irreversibilityDeclared + flagged before commitNot modeledNot modeledManual
LLM in the rollback pathNever — deterministic (anti prompt-injection)Your code decidesOften the LLM decidesDepends
Tamper-evident auditHash-chained, anyone verifiesWorkflow history (trust the server)App logsManual
Setup costnpm i -g chronomcp, transparentSignificant (SDK, workers, infra)Adopt the frameworkPer-tool code
Best forGuarding any agent's tool calls with human oversightComplex, long-running business workflows you ownBuilding the agent's reasoning graphSmall, fully-controlled cases

green = strength · amber = partial/depends · red = not designed for it. This compares categories; every tool listed is excellent at what it was built for.

When to use which

Reach for ChronoMCP when…

  • an AI agent calls tools that can hurt (delete data, send email, move money);
  • you want a human in the loop — approval, an impact preview, or a Slack/dashboard gate;
  • you need an honest record of what could and couldn't be undone;
  • you want it without rewriting your agent — it drops in front of any MCP server.

Reach for an orchestrator when…

  • you're building a complex, long-running workflow you fully own and control;
  • you need durable execution, retries, timers and state across days;
  • the compensation logic is deep in your own business code, not at a tool boundary;
  • the workflow isn't necessarily driven by an MCP agent at all.

The honest verdict: ChronoMCP is the guardrail at the boundary between an AI agent and the real world; an orchestrator is the engine that runs a workflow you built. Many teams run both — the orchestrator drives the durable business process, and ChronoMCP guards the individual, risky tool calls the agent makes, with a human able to say "no" before anything irreversible happens.

Why "honest about irreversibility" matters here

Most "undo" stories quietly break at the edges: a refund is a new transaction, not a reversal; a sent email cannot be un-sent. ChronoMCP makes irreversibility a first-class, declared property (via the open mcp-compensate extension) and surfaces it before you approve — not as a surprise after a failure. That is the difference between a guard you can trust and a rollback that lies.

See how it works →