Dagpipe
Run multi-step LLM workflows that survive crashes and resume without standing up a workflow database.
Overview
dagpipe is a MCP server for the Build phase that exposes crash-proof LLM pipelines that resume from failure without a database.
What is this MCP server?
- Crash-proof LLM pipelines with resume-from-failure semantics
- No separate database required for pipeline checkpointing (per server description)
- PyPI package dagpipe-core v0.2.3 with stdio MCP transport
- Suited for long agent chains that must not restart from step one after errors
- Open-source repo at github.com/devilsfave/dagpipe
- Registry version 0.2.3
- PyPI identifier: dagpipe-core
- Transport: stdio
Community signal: 3 GitHub stars.
What problem does it solve?
Long LLM pipeline runs die on one bad step and force you to rebuild state or add heavy persistence just to retry.
Who is it for?
Solo builders running multi-step LLM jobs (enrichment, batch codegen, eval loops) who want lightweight durability.
Skip if: Simple one-shot completions or teams that already standardize on Airflow/temporal with full ops tooling.
What do I get? / Deliverables
After wiring dagpipe-core via MCP, agents can run pipelines that pick up after failures instead of restarting from scratch.
- MCP-connected dagpipe runtime for agent-driven pipelines
- Resumable execution path after step failures
- Less bespoke retry/state code in agent automation scripts
Recommended MCP Servers
Journey fit
Dagpipe targets how you orchestrate LLM steps during product build; resilient pipelines are agent-tooling infrastructure, with recovery benefits when you operate those jobs in production. Agent-tooling is the primary shelf because the server advertises crash-proof LLM pipelines—core builder automation, not a one-off integration.
How it compares
LLM pipeline runtime over MCP, not a hosted workflow UI or generic chat skill.
Common Questions / FAQ
Who is dagpipe for?
Developers shipping agent-heavy automations who need dependable multi-step LLM execution without managing a pipeline database.
When should I use dagpipe?
Use it while building agent workflows that must survive restarts, rate limits, or transient API errors mid-pipeline.
How do I add dagpipe to my agent?
Install the dagpipe-core PyPI package (0.2.3), configure the stdio MCP server in your client, and call the exposed tools from your agent session.