Skip to content
Agents tracked: 501 Releases (7d): 777 Agent status: no reported issuesSince last limit reset: Claude 10d · Codex 8d For AI agents: Markdown Updated Nov 21, 2026
Case study SpotifyResults as reported by Spotify

How Spotify used its Honk background coding agent to open 240 data-pipeline migration PRs

Spotify used Honk, its internal background coding agent built on Claude Code capabilities, to move downstream consumers off two deprecated datasets. The agent opened 240 migration pull requests, and the owning teams reviewed and merged them.

The problem

Two heavily used datasets were being deprecated. About 1,800 downstream pipelines depended on them, spread across BigQuery Runner, dbt and Scio, and they had to move within six months. Doing it by hand would have taken a lot of engineering time.

How they did it

  1. Found affected repositories with Backstage's lineage and Codesearch plugins.
  2. Queued the target repositories for change through Fleetshift, Backstage's plugin for large-scale code changes.
  3. Wrote detailed context files with field-mapping tables, since Honk could only see the context written for it.
  4. Left out the highly variable Scio pipelines and focused on the more standard dbt and BigQuery Runner code.
  5. For cases that needed judgment, told the agent to leave fields unchanged and add comments linking to human migration guides.
  6. Owning teams tested and merged the PRs themselves, because these repos rarely had build-time unit tests.

Results as reported by Spotify

  • As reported by Spotify, the team 'successfully rolled out 240 automated migration PRs using Fleetshift'.
  • The PRs covered dbt and BigQuery Runner pipelines; Scio pipelines were left out, and Spotify does not say how many of the ~1,800 affected pipelines were migrated this way.
  • Spotify estimated the work would have taken 'around 10 engineering weeks of effort' to complete by hand.

These figures are as reported by Spotify in Spotify Engineering blog ↗ (Apr 22, 2026). AgentGid did not measure or independently verify them.

Takeaway. Before you hand a migration to an agent, put the mappings in explicit tables and say how it should flag cases it can't decide.
AgentGid's take

This suits a platform team with a large fleet of similar repos, a lineage tool, and engineers willing to write detailed mapping context. Honk is Spotify's internal tool, so you'd have to rebuild that layer yourself. Also note the 10-week saving is Spotify's own estimate, and owning teams still reviewed every PR. For a cheaper start, Pi is free (OSS), though you pay your model provider.

The agent used here

Similar use cases

Guides