← All implementations

Replit · Orchestration system

Manager agent (agent-of-agents)

An internal agent-of-agents stack where every employee gets a manager agent that spawns multiple agents for verifiable work and escalates judgment to humans.

1

Approach type
Orchestration system
Work
Coding, Code review, Support, Research, Data
Human involvement
Drafts reviewed
Invocation
Interactive
Interfaces
Slack
Deployment stage
Scaled
Evidence strength
Detailed primary
Entry reviewed

How it works

The workflow the sources report for this implementation.

One human gives an objective; the manager spawns parallel agents for verifiable work and escalates judgment

1

Where people stay involved

  • objective → verifiable multi-agent work productWork product review · Level 3

Level 3 for objective → verifiable multi-agent work product; human attention boundary: work-product-review.

1

Observation date
2026

Implementation details

Sandbox

microVMs and remote filesystems behind access policies, token proxies, audit logging, and a ZeroTrust network

1

Harness

Fleet/loop orchestration: a manager agent launches parallel agents for verifiable work and escalates judgment

1

Model

Not specified

1

Interfaces

slack

1

Tool access

Investigates incidents, reviews PRs, answers questions, analyzes company data, triages support, researches sales accounts, improves Replit Agent itself

1

Context management

Manager agent coordinates parallel sub-agents and routes results

1

Reported results and limitations

The catalog records what the sources report, with the scope and the denominator of every figure. A qualification below limits the figure it sits under.

Reported metrics

Headline claim

2.9x code output for a consistent author cohort; review latency, PR reversions, and incident trends reported flat

1

Reported by
Replit
Scope
Code output for a consistent author cohort across early January to late June; separate from company-wide hiring effects
Denominator
Same cohort of authors before and after
Method
Comparison of contributed code for a consistent author cohort; raw company-wide lines of code increased 5.8x
Key observation

2.9x code output for a consistent author cohort from early January to late June

1

Reported by
Replit
Scope
Code output for a consistent author cohort across early January to late June; separate from company-wide hiring effects
Denominator
Same cohort of authors before and after
Method
Comparison of contributed code for a consistent author cohort; raw company-wide lines of code increased 5.8x
Key observation

No corresponding deterioration in review/reversion/incident metrics

1

The source does not report the denominator of this figure.

Reported by
Replit
Scope
Company code review latency, PR reversion rates, and incidents opened during increased code output
Method
Company comparison of review latency, PR reversion rates, and incident trends

Lessons and interpretation

Give every employee a manager agent that spawns sub-agents; verifiable work parallelizes, judgment escalates to humans

1

Track outcome metrics (reverts, incidents), not activity; output can scale without quality regressions

1

Sources and research details

Citations link to the original publisher. Each source also keeps a preserved copy in the repository, so a changed or removed page stays checkable.

  1. The Self-Driving Companyhttps://ld.replit.com/blog/self-driving-companyEngineering blog · First party · Last source verification: 2026-08-31
    Preserved copy ↗
Research details for every claim on this page
  1. Summary
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  2. Headline claim
    Statement type
    Metric
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
    Reported by
    Replit
    Scope
    Code output for a consistent author cohort across early January to late June; separate from company-wide hiring effects
    Denominator
    Same cohort of authors before and after
    Method
    Comparison of contributed code for a consistent author cohort; raw company-wide lines of code increased 5.8x
    Observation date
    Not reported
  3. Sandbox
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  4. Harness
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  5. Model
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  6. Interfaces
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  7. Tool access
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  8. Context management
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  9. Supporting component
    Statement type
    Fact
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
  10. Key observation
    Statement type
    Metric
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
    Reported by
    Replit
    Scope
    Code output for a consistent author cohort across early January to late June; separate from company-wide hiring effects
    Denominator
    Same cohort of authors before and after
    Method
    Comparison of contributed code for a consistent author cohort; raw company-wide lines of code increased 5.8x
    Observation date
    Not reported
  11. Key observation
    Statement type
    Metric
    Provenance
    Reported
    Confidence
    High
    Confidence reason
    A linked first-party source states the claim.
    Reported by
    Replit
    Scope
    Company code review latency, PR reversion rates, and incidents opened during increased code output
    Denominator
    Not reported
    Method
    Company comparison of review latency, PR reversion rates, and incident trends
    Observation date
    Not reported
  12. Lesson
    Statement type
    Inference
    Provenance
    Catalog judgment
    Confidence
    Medium
    Confidence reason
    The catalog derives this observation from the linked sources.
  13. Lesson
    Statement type
    Inference
    Provenance
    Catalog judgment
    Confidence
    Medium
    Confidence reason
    The catalog derives this observation from the linked sources.
  14. Operating model assessment
    Statement type
    Inference
    Provenance
    Catalog judgment
    Confidence
    High
    Confidence reason
    The source describes autonomous parallel execution followed by human judgment on the resulting work.
    Observation date
    2026

The catalog records no related implementation for this entry yet.

Back to the directory →