Iterward MCP server
Commit Oct 22, 2026Provides a deterministic control plane for agentic workflows by managing routing, retries, approvals, and stop conditions via loop files.
Also known as push-based/iterward
- Stars
- 5
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Oct 22, 2026
- License
- MIT
What is Iterward MCP?
Iterward provides a deterministic control plane for agentic workflows by managing routing, retries, approvals, and stop conditions via loop files.
Summary generated from the project's README
About Iterward MCP server
Managing agentic engineering workflows through deterministic protocols is what Iterward enables for developers. Instead of relying on prompts to handle complex logic, this control plane uses loop files to manage routing, retries, human approvals, and stop conditions. This approach ensures that agent execution follows strict transitions while recording every human interaction.
Users can integrate this server into Claude Code by adding it to a local .mcp.json configuration file. The server requires Node.js 22.18 or later and operates via the npx command. It provides specific tools like build-loop, which allows an agent to interview a user to design and write a functional loop file. By moving decision-making from the LLM to a structured loop file, developers can test their workflows using node:test and maintain precise control over how agents respond to failures or verification steps.
What you can do with it
- 1Automate workflow logicMove routing and retry logic out of LLM prompts and into a testable loop file.
- 2Implement human-in-the-loop gatesPause an agentic run until a person provides manual approval.
- 3Design new agent protocolsUse the build-loop tool to have an agent interview you and generate a working loop configuration.
- 4Prevent infinite agent loopsSet specific limits to stop a run after a certain number of attempts.
Written from the project's README. Every use case is checked against a quote from it.
Automated checks
Ran Nov 15, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablepush-based/iterward is public and reachablegithub.com/push-based/iterward ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-10-22github.com/push-based/iterward ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHub
LLMs do the work. Iterward decides what happens next.
Iterward is a deterministic control plane for agentic engineering workflows. It moves routing, retries, approvals, failures, and stop conditions out of prompts and into a loop file and code that you can test. Agents own execution, and Iterward owns protocol: a run moves only along the transitions its loop file declares, and Iterward records every human move.
Iterward closes the loop around your agents. In a closed-loop agent workflow, each result decides the next step: a failed verification sends the work back to the builder, a gate holds the run until a human approves, and a limit parks the run after too many attempts. Iterward makes those decisions from the loop file, so your agents don't have to.
Get started
Iterward needs Node.js 22.18 or later and has no npm dependencies. To copy the default loop into your repository and start a run, run these commands from the repository root:
npx @push-based/iterward init
npx @push-based/iterward new --config loops/default/default.loop.json --run runs/example/run.json --subject 'the settings page'
npx @push-based/iterward next --run runs/example/run.json
next prints the first dispatch: the message for the planner agent. To install Iterward as a development dependency instead, see Install.
Use it from an agent
The MCP server exposes the CLI's commands as tools. To register it with Claude Code, add it to the .mcp.json file in your repository root:
{
"mcpServers": {
"iterward": { "command": "npx", "args": ["-y", "@push-based/iterward@latest", "mcp"] }
}
}
To keep a run on the version that started it, see Register the server. The server's build-loop tool gives an agent the build-loop skill, which interviews you about a workflow and writes a loop file that compiles.
Documentation
The following pages cover the rest:
- Iterward: the problem, how Iterward works, a loop in code, and what the engine enforces.
- Run the default loop by hand: play every agent and take a run to
done. - Configuration: the loop file, its hooks, and the standard message.
- CLI reference: the commands, their flags, and the exit codes.
- MCP server: drive a run and design a loop through MCP tool calls.
- Protocol rules: the promise, the terms, and the rules.
- Testing: test a loop with
node:test.
Develop
The repository is an Nx workspace with one package, packages/iterward. To build and test it, install the workspace, and run the targets through Nx:
npm install
npx nx run-many -t build test
Every commit follows Conventional Commits. A commit-msg hook, which npm install sets up, checks each message.
License
Iterward is available under the MIT License.
Iterward MCP questions
What does Iterward do for agentic workflows?
It acts as a deterministic control plane that moves routing, retries, approvals, and stop conditions out of prompts and into testable loop files.
How can I use Iterward with Claude Code?
You can register it by adding the server configuration to your .mcp.json file in your repository root using npx.
What are the system requirements for Iterward?
Iterward requires Node.js version 22.18 or later.
Can Iterward help design a workflow?
Yes, the build-loop tool allows an agent to interview you about a workflow and generate a compiling loop file.