Flux
Flux is DoorDash’s cloud platform for engineering agents. It supplies isolated sandboxes, a governed MCP gateway, reusable playbooks and invocation surfaces. DoorDash’s analytics AI Marketplace is a separate documented platform.
- Company
- DoorDash
- Approach type
- Platform
- Work supported
- Code review, Coding, CI triage, On-call, Maintenance
- Interfaces
- Slack, Github, Scheduled, Cli, Skill
- Deployment stage
- Scaled
- Evidence strength
- Detailed primary
- Entry reviewed
Capabilities and architecture
- Model
- Not reportedNo model or provider is named for either platform. The sources describe modular primitives that allow the best third-party tool or an in-house build for each job, with multiple supported coding agent harnesses.
- Harness
- Separate analytics AI Marketplace context (not a Flux capability): Maturity model: deterministic workflows → ReAct agents → hierarchical deep agents → experimental swarms[Source 2]The maturity model is the analytics platform's progression. Flux supports multiple coding agent harnesses inside its sandboxes.
- Sandbox
- Firecracker microVMs; <5s p95 end-to-end setup (boot, clone repos, install tools, configure harness)[Source 3]Firecracker microVMs with a sub-five-second p95 end-to-end setup objective; a Flux capability.
- Tool access
- Flux Agent Gateway brokers playbook-declared tools with scoped, logged permissions. The separately documented analytics AI Marketplace uses LangGraph and explores A2A.[Source 2, Source 3]Flux Agent Gateway with playbook-scoped logged permissions; LangGraph orchestration and prospective A2A on the analytics platform.
- Knowledge
- Flux playbooks package DoorDash-specific context. The separate analytics AI Marketplace hosts DataExplorer.[Source 2, Source 3]
- Context management
- Separate analytics AI Marketplace context (not a Flux capability): Hybrid retrieval: BM25 + dense semantic + reciprocal-rank fusion → RAG; schema-aware SQL with EXPLAIN validation[Source 2]Hybrid retrieval and schema-aware SQL generation; an analytics-platform capability.
- Credentials
- Scoped per playbook; brokered through the gateway, never on the laptop; provenance on every action[Source 3]Scoped per playbook and brokered through the gateway rather than the laptop; a Flux capability with provenance on every action.
- Interfaces
Flux invocation surfaces are Slack, GitHub, cron, CLI, and skill; the analytics platform adds Cursor. The analytics conversational web UI is not claimed.
- slack
- github
- scheduled
- cli
- skill
Mechanisms
Isolated Firecracker microVM with repos, tools, secrets, runtime deps[Source 3]
Governed, audited access to CI, observability, issue trackers, deploy, code search[Source 3]
YAML unit of agentic work: task, inputs, skills, tools, permissions, validation, outputs[Source 3]
Separate analytics AI Marketplace context (not a Flux capability): Identifies schemas, generates grounded SQL, validates via EXPLAIN before execution[Source 2]
Separate analytics AI Marketplace context (not a Flux capability): Deterministic workflows and single agents are in use; deep agents are being developed and tested; swarms remain research[Source 2]
Documented uses
Business question to grounded, validated answer on the analytics platform.
Separate analytics AI Marketplace context (not a Flux capability): From a business question, DataExplorer identifies candidate tables through DescribeTable with cached column examples, generates starter SQL grounded in the schema, and EXPLAIN validation with autocorrection precedes execution[Source 2]
Access and controls
See documented credential and access boundaries in the architecture above. Scoped human-review assessments for individual uses remain in research details.
Reliability and validation
Reported: Answer a grounded data question, Lesson
Adoption and operating evidence
These results come from the sources. Notes below each result explain what it covers and any reported limits.
130,000 engineering tasks automated in one month[Source 1]
- Reported by
- DoorDash
- Scope
- Engineering tasks automated in one reported month; calendar measurement month unspecified
The source does not report the denominator of this figure.
Observed on 11 August 2026
About this result
- Category
- Adoption output
- Evidence type
- Reported measurement
- What it covers
- Engineering tasks automated on Flux in one reported month; the calendar month is unspecified
25,000+ automated code reviews per week[Source 1]
- Reported by
- DoorDash
- Scope
- Weekly automated code reviews powered by Flux
The source does not report the denominator of this figure.
Observed on 11 August 2026
About this result
- Category
- Adoption output
- Evidence type
- Reported measurement
- What it covers
- Weekly automated code reviews powered by Flux; the reviewer agent itself is the separate doordash-code-review record
300+ playbooks; 10,000+ invocations per week[Source 3]
- Reported by
- DoorDash
- Scope
- Unique playbooks and weekly invocations on Flux
The source does not report the denominator of this figure.
About this result
- Category
- Adoption output
- Evidence type
- Reported measurement
- What it covers
- Unique Flux playbooks and weekly playbook invocations
Lessons
DoorDash first used Flux for code review, then expanded into CI triage, on-call work, maintenance, and ticket-driven development.[Source 3]
Reported opinion: DoorDash credits a move from private run channels to public Slack threads with helping teams learn to use Flux.[Source 3]
DoorDash ran workshops and hackathons to help teams turn recurring operational tasks into Flux playbooks.[Source 3]
DoorDash reports deterministic workflows and single agents in use, deep agents under development, and swarms as research on its analytical platform.[Source 2]
DoorDash's analytical platform uses SQL linting and EXPLAIN checks, with separate model judges for response quality.[Source 2]
DoorDash describes recording the queries, documents, and agent interactions behind an analytical answer.[Source 2]