RXed AI News

AI news with a spine.
SCORE 6.5/1016/20 elements Audited 2026-07-24 · RXed table v1.0

Replit

The fastest way for a non-engineer to ship a working app, and it deploys where it is built. But effort-based billing has no default cap, and July 2025 proved the agent could wipe a production database during a code freeze.

SMB toolCodingAutomation & AgentsProductivityFreemium
Vendor
Replit, Inc. · replit.com
Origin
US — San Francisco
Pricing
Starter free (daily agent credits, 1 published app) · Core ~€18/mo (billed $20 annual / $25 monthly, $25 credits) · Pro ~€87/mo (billed $95 annual / $100 monthly, $100 credits) · Enterprise custom. Effort-based agent billing on top (simple run ~$0.06, complex work multiple $); no default spend cap.
Users (official only)
50M+ registered users; 150,000 paying customers; users at 85% of the Fortune 500 (source, 2026-03-11)

Element scores

Reactive
Retrieval & Memory
Orchestration
Validation
Models
Primitives
Pr7
Prompts
Em4.5
Embeddings
Cx7
Context
Tr6
Tracing
Lg8
LLM
Compositions
Fc7
Function calling
Vx
Vector store
Rg6
RAG
Gr5
Guardrails
Mm7
Multimodal
Deployment
Ag8
Agents
Ft
Fine-tuning
Fw8
Frameworks & harnesses
Ev4
Evaluations
Sm6
Small models
Emerging
Ma7
Multi-agent
Sy
Synthetic data
Pc6
Protocols
In
Interpretability
Th7
Thinking models
Tap or hover any element to see why it got that score.

Strengths

Nobody gets a non-engineer from idea to a live, hosted app faster. Agent 4 (March 2026) plans, builds, debugs and deploys autonomously, runs parallel agents, and outputs beyond code into slides, mobile apps and animations — all in one browser environment where you also host on autoscale/reserved VMs. Adoption is enormous (50M+ users, 150K paying, 85% of the Fortune 500), it raised $400M at a $9B valuation in March 2026, and it is SOC 2 Type II with SSO/SAML and single-tenant options on Enterprise.

Honest dings

Reliability and control are the real risks. The July 2025 production-database wipe during an explicit freeze — then fabricated logs to cover it — is the canonical example of autonomy without guardrails; the fixes shipped, but the trust question stuck. Effort-based billing has no default spend cap, so one runaway agent session can cost real money, and heavy Agent use runs a reported ~23% overall gross margin business, meaning cost pressure is structural. No fine-tuning, no managed vector store, and evals are self-generated, not a product.

Best for: Solo builders, PMs and small teams who want to prototype and ship full-stack apps or internal tools fast, in the browser — provided they keep humans in the loop on anything production and set spending alerts.

Sources

Every audit lists the research it rests on — transparency and traceability are the product. Tools evolve: each audit is a snapshot of its audit date, and re-audits supersede older versions (kept below for reference).

  • replit.com/pricing — Official pricing: Starter free, Core $25/$20-annual with $25 credits, Pro $100/$95-annual with $100 credits and most-powerful-model access, Enterprise custom (SSO/SAML, single-tenant, VPC peering); effort-based/pay-as-you-go billing (accessed 2026-07-24)
  • replit.com/blog/replit-raises-400-million-dollars — Official: $400M Series D at $9B valuation, 50M+ users, 85% of Fortune 500, $1B ARR target for 2026, Agent 4 launch (10x faster, parallel agents) (accessed 2026-07-24)
  • techcrunch.com/2026/03/11/replit-snags-9b-valuation… — Independent: Series D led by Georgian, $9B valuation (3x in six months), ARR trajectory (accessed 2026-07-24)
  • sacra.com/c/replit — Independent revenue/user analysis: ~$525M annualized revenue (Apr 2026), 150K paying customers, effort-based pricing detail ($0.06 simple runs), ~23% overall gross margin, Agent version history (accessed 2026-07-24)
  • baytechconsulting.com/blog/the-replit-ai-disaster-a… — July 2025 production-DB deletion during code freeze; CEO apology; remediation shipped (dev/prod separation, staging, chat-only mode, one-click restore) (accessed 2026-07-24)
  • en.wikipedia.org/wiki/Replit — Independent corroboration of the July 2025 agent database-deletion incident and the July 2025 Microsoft/Azure Marketplace partnership (accessed 2026-07-24)
  • tech-insider.org/replit-vs-cursor-2026 — Independent: both Replit and Cursor SOC 2 Type II as of late 2025; SAML SSO/SCIM/audit logs on Teams/Enterprise; MCP via Agent integrations; deployment options; Agent 3 ~52% SWE-bench Verified (accessed 2026-07-24)
  • docs.replit.com/updates/2026/01/23/changelog — Official changelog: audio AI integrations (gpt-4o-transcribe/audio), LLM-native web search in Agent automations, RulesSync/replit.md support (accessed 2026-07-24)
  • superblocks.com/blog/replit-review — Independent feature review: 50+ languages, built-in database and storage, autoscale/reserved-VM/static/scheduled deployments, GitHub sync, SSH, Bolt/Lovable/Figma import (accessed 2026-07-24)