Key takeaways
- Model flexibility is table stakes, but automation and integration depth now differentiate leaders
- Four platforms now offer both scheduled and event-triggered automations: Tembo, Factory, Ona, and Devin — up from two in March 2026
- Tembo is uniquely both model-agnostic AND CLI-agnostic — orchestrating any coding agent without lock-in
- Surface coverage varies wildly: Cursor owns IDE, OpenCode owns CLI, most agents are web-only
- Database integrations (PostgreSQL, Supabase, AWS RDS) are unique to Tembo — enabling schema-aware agents
FAQ
What is a model-agnostic coding platform?
A model-agnostic coding platform works with multiple AI providers (Anthropic, OpenAI, Google) rather than being locked to a single model vendor, giving teams flexibility as the AI landscape evolves.
Which platform has the best automations?
Four platforms now offer both scheduled and event-triggered automations: Tembo (11 integrations), Factory (4-5), Ona (enterprise-grade), and Devin (self-scheduling agents). Blocks offers event-triggered automations.
What's the difference between coding agents and agent orchestrators?
Coding agents (Devin, Claude Code) execute work directly. Orchestrators (Tembo, 8090) coordinate multiple agents and integrate with team workflows like Jira, Linear, Slack, and Sentry.
Which platforms support local development?
Cursor, OpenCode, Tembo, Amp, Background Agents, and Conductor support local development. Devin, Factory, 8090, cto.new, and Ona are cloud-only.
Executive Summary
The AI coding tool market is rapidly fragmenting — no single model or agent will win. Model-agnostic platforms that work across providers are becoming essential infrastructure for teams hedging their bets.
This report evaluates 14 platforms across seven key dimensions: model flexibility, CLI/agent flexibility, automation capabilities, integration depth, surface coverage, local development support, and enterprise readiness.
Key Findings:
- Model flexibility is table stakes — 10 of 14 platforms support multiple AI providers, but this alone no longer differentiates
- Automation depth separates leaders — Four platforms now offer both scheduled and event-triggered automations: Tembo, Factory, Ona, and Devin. Blocks adds event-triggered automations. The rest remain interactive-only
- Integration breadth matters — Tembo leads with 11 native integrations (GitHub, GitLab, Bitbucket, Linear, Jira, Sentry, PostgreSQL, Supabase, AWS RDS, Slack, Raycast); competitors average 2-4
- CLI-agnostic is the new model-agnostic — Tembo uniquely orchestrates any coding agent (Claude Code, Codex, Amp, OpenCode); competitors lock you to their agent
- Surface coverage varies wildly — Cursor dominates IDE, OpenCode dominates CLI, but most agents (Devin, Ona, 8090) are web-only with no local option
- Database integrations are rare — Tembo is the only platform with native PostgreSQL, Supabase, and AWS RDS support for schema-aware agent work
Strategic Planning Assumptions:
- By 2027, teams will evaluate platforms primarily on automation and integration capabilities, not just model support
- By 2028, orchestration layers with deep integrations become mandatory for Fortune 1000 AI coding deployments
- By 2029, platforms without scheduled automations will be relegated to individual developer tools, not team infrastructure
Market Definition
Model-agnostic agentic engineering platforms are AI-powered development tools that support multiple language model providers (Anthropic, OpenAI, Google, etc.) or multiple coding agents, avoiding vendor lock-in while delivering autonomous or semi-autonomous code generation capabilities.
Inclusion Criteria:
- Supports at least two major LLM providers OR multiple coding agents
- Provides code generation, editing, or orchestration capabilities
- Available as commercial product or mature open-source project
- Active development and user base as of February 2026
Exclusion Criteria:
- Single-model-only tools (Claude Code CLI, Codex CLI)
- Pure autocomplete/suggestion tools without agentic capabilities
- Research projects or early prototypes without user adoption
- IDE plugins that only support one model provider
Comparison Matrix
| Vendor | Category | Model/Agent Support | Automations | Integrations | Local Dev | Pricing |
|---|---|---|---|---|---|---|
| Tembo | Orchestrator | 5+ agents (CLI-agnostic) | Cron + Events | 11 native | — | $60-200/mo |
| Factory | Agent | Multi-model (3+) | Cron + Events | 4-5 | — | Token-based |
| Cursor | IDE | Multi-model (5+) | — | 3 (git only) | ✅ | $60/mo |
| OpenCode | Agent | 75+ providers | — | 3 (git only) | ✅ | OSS (free) |
| Devin | Agent | Custom models | Cron + Events | 2-3 | — | ~$500/seat/mo |
| 8090 | Orchestrator | Agent-agnostic | Limited | 2-3 | — | $200/seat/mo |
| Amp | Agent | Multi-model (3+) | — | 1-2 | ✅ | Free + usage |
| Conductor | Orchestrator | 2 agents | — | 1 | ✅ | $22M Series A |
| cto.new | Agent | Multi-model (3+) | — | 1-2 | — | Free |
| Ona | Agent + Env | Not disclosed | Cron + Events | 2-3 | — | Enterprise |
| Background Agents | Infrastructure | Via OpenCode | — | 1 | ✅ | OSS (free) |
| Blocks | Orchestrator | 7+ agents (agent-agnostic) | Events | 8+ (GitHub, GitLab, Bitbucket, Slack, Linear, Jira, Sentry, Notion) | — | Free (beta) |
| Niteshift | Agent | Multi-model (2+) | — | 3 (Slack, Linear, Figma) | — | TBA (waitlist) |
| Replicas | Agent | Claude Code + Codex | Events only | 4 (GitHub, Slack, Linear, API) | — | Unknown |
| 20x | Agent | Multi-agent (Claude Code, OpenCode, Codex) | Skills | 3 (Linear, HubSpot, GitHub) | Electron | Open source (peakflo/20x) |
| Entire | Context Capture | Agent-agnostic | — | 1 (git) | ✅ | $60M seed (free/MIT) |
Sorted by automation + integration capabilities
Product Profiles
8090
Full-SDLC orchestration platform capturing requirements, architecture, and context upstream.
| Quick Reference | |
|---|---|
| Website | 8090.ai |
| Founded | Unknown |
| Funding | Not disclosed |
Overview
8090 takes a contrarian approach: while competitors race to generate code faster, Software Factory focuses on what comes before code — requirements, architecture decisions, and planning context. The platform creates a knowledge graph that feeds AI agents better inputs.
Strengths
- Full-SDLC scope — Covers requirements through validation, not just code generation
- Context-first architecture — Structured knowledge graph gives AI agents better inputs
- Enterprise modernization play — Reverse engineering agents for legacy codebases
- Differentiated positioning — Complements rather than competes with coding agents
Cautions
- Complexity — Four modules may overwhelm teams wanting simple solutions
- Pricing opacity — Token-based usage on top of seat fees complicates cost prediction
- Limited visibility — Minimal public traction and no disclosed funding
Key Stats
| Metric | Value |
|---|---|
| Pricing | $200/seat/mo + tokens |
| Model Support | Agent-agnostic (external) |
| Enterprise Tier | Yes ($1M+/year custom) |
Amp
Independent AI research lab building a multi-model terminal-first coding agent.
| Quick Reference | |
|---|---|
| Website | ampcode.com |
| Founded | 2025 (spun out from Sourcegraph Dec 2025) |
| Funding | Profitable, independent |
Overview
Amp spun out from Sourcegraph in December 2025 to become Amp, Inc. — an independent AI research lab focused on the frontier of AI-assisted development. February 2026: Amp made a bold strategic pivot — killed its IDE editor extension entirely ("The Coding Agent Is Dead") and went all-in on pure CLI/terminal. Now ad-free on the free tier, pay-as-you-go with no markup. Uses Claude Opus 4.5 in smart mode, GPT-5 for deep reasoning, and fast models for quick operations — automatically routing to the best model per task. An interesting positioning bet that abandons the IDE war entirely.
Strengths
- Multi-model routing — Automatically selects optimal model for each task type
- Transparent pricing — No markup on model costs; $10/day free tier
- Independent and profitable — Sustainable without VC pressure
- Opinionated quality — Ruthlessly removes features that don't work (killed Tab, TODOs, Fork)
Cautions
- Terminal-only — Killed IDE extension entirely; alienates GUI-preferring developers
- Breaking changes — "No backcompat, no legacy features" policy
- Contrarian bet — Abandoning IDE market may limit addressable market
- Limited integrations — No Jira, Linear, Sentry, or database integrations
Key Stats
| Metric | Value |
|---|---|
| Pricing | Free ($10/day grant) + usage |
| Model Support | Claude Opus 4.5, GPT-5, Gemini 3 |
| Status | Independent company (Dec 2025) |
Background Agents
Open-source background agent system using Modal sandboxes and OpenCode.
| Quick Reference | |
|---|---|
| Website | GitHub |
| Founded | 2026 |
| Funding | None (OSS) |
Overview
Background Agents (Open-Inspect) is an open-source implementation inspired by Ramp's internal Inspect tool. Using Modal for cloud sandboxes and OpenCode as the coding agent runtime, it enables asynchronous agent work while developers focus elsewhere. Single-tenant only, designed for internal team use.
Strengths
- Fully open-source — MIT licensed, inspect and modify everything
- Modern architecture — Cloudflare edge + Modal sandboxes
- Multiplayer support — Collaboration features rare in OSS agent tools
- Active development — 537 GitHub stars in first 2 weeks
Cautions
- Single-tenant only — Not suitable for multi-org or commercial deployment
- Infrastructure dependencies — Requires Modal (vendor lock-in for compute)
- No enterprise features — Missing Jira, signed commits, audit logs, BYOK
Key Stats
| Metric | Value |
|---|---|
| Pricing | Free (OSS) + infra costs |
| Model Support | Via OpenCode (configurable) |
| GitHub Stars | 537 |
Blocks
Agent-agnostic orchestration platform with event-driven automations, enterprise security, and deep integration coverage.
| Quick Reference | |
|---|---|
| Website | blocks.team |
| Founded | Unknown |
| Funding | Not disclosed |
| Team | HPC orchestration background (Stanford, Brookhaven National Lab, IBM, Google, UC Berkeley) |
Overview
Blocks has emerged as one of the most direct competitors in the orchestration space. The platform supports 7+ coding agents (Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Kimi Code, Sisyphus), offers event-triggered automations (GitHub Actions failures, PR events, GitLab MR, Bitbucket push), and matches enterprise requirements with SSO, audit logging, RBAC, self-hosting, and zero data retention. Integration coverage has expanded significantly to include GitHub, GitLab, Bitbucket, Slack, Linear, Jira, Sentry, Postgres, and Notion.
Strengths
- Agent-agnostic — 7+ coding agents including Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Kimi Code, and Sisyphus
- Deep integrations — GitHub, GitLab, Bitbucket, Slack, Linear, Jira, Sentry, Postgres, Notion — rivaling Tembo's coverage
- Event-driven automations — Triggers on GitHub Actions failures, PR events, GitLab merge requests, Bitbucket push
- Enterprise-ready — SSO, audit logging, RBAC, self-hosting, zero data retention
- Strong team pedigree — HPC orchestration background from Stanford, Brookhaven National Lab, IBM, Google, UC Berkeley
- Active thought leadership — Blog posts on "Turn Coding Agents Into CI Gates" (Mar 27) and "The Agentic Cost Cliff" (Mar 22)
Cautions
- No scheduled automations — Event-triggered only; no cron-style scheduling for recurring tasks
- Beta stage — Pre-GA with limited quotas; pricing unknown
- Cloud-only — No local development or CLI support
- Unknown funding — No disclosed funding or traction metrics
Key Stats
| Metric | Value |
|---|---|
| Pricing | Free (beta), TBA for GA |
| Agent Support | Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Kimi Code, Sisyphus |
| Integrations | GitHub, GitLab, Bitbucket, Slack, Linear, Jira, Sentry, Postgres, Notion |
| Enterprise | SSO, audit logging, RBAC, self-hosting, zero data retention |
Conductor
Mac desktop app for running parallel Claude Code and Codex agents in isolated worktrees.
| Quick Reference | |
|---|---|
| Website | conductor.build |
| Founded | 2024 |
| Funding | $22M Series A |
Overview
Conductor is a Mac-native application for running multiple coding agents in parallel. Using git worktrees for isolation, it enables concurrent agent work without merge conflicts. Now backed by a $22M Series A and trusted by engineering teams at Linear, Vercel, Notion, and Ramp, Conductor has proven product-market fit for Mac-native parallel agent work.
Strengths
- True parallelism — Multiple agents working simultaneously on same repo
- No additional costs — Uses existing Claude/Codex subscriptions
- Clean isolation — Git worktrees prevent agents from conflicting
- Native performance — Mac-native app, not Electron
- Strong traction — Trusted by Linear, Vercel, Notion, Ramp
- Well-funded — $22M Series A provides significant runway
Cautions
- Mac-only — No Windows, Linux, or cloud support
- Limited agent support — Only Claude Code and Codex
- No enterprise features — No team management, Jira, or signed commits
- Individual-focused — No collaboration features for teams
Key Stats
| Metric | Value |
|---|---|
| Pricing | Unknown (subscription) |
| Funding | $22M Series A |
| Agent Support | Claude Code, Codex |
| Customers | Linear, Vercel, Notion, Ramp |
| Platforms | macOS only |
cto.new
Completely free AI coding agent with frontier models, funded by data monetization.
| Quick Reference | |
|---|---|
| Website | cto.new |
| Founded | 2025 |
| Funding | $5.7M Seed |
Overview
cto.new provides free access to GPT-5.2, Claude Sonnet 4.5, and Gemini 3 Pro Preview — models that competitors charge $20-200/month for. The platform monetizes through data collection and ads, processing 300B+ tokens and thousands of daily PRs across 10,000+ codebases since October 2025 launch.
Strengths
- Zero cost barrier — No credit card, API key, or subscription required
- Frontier model access — Same models competitors charge premium prices for
- Proven scale — 300B+ tokens, 10,000+ codebases, thousands of daily PRs
- IDE integration — MCP server connects to Cursor for hybrid use
Cautions
- Sustainability questions — Free model requires significant token subsidies
- Privacy trade-offs — Data collection is part of monetization strategy
- New company risk — Limited operating history since October 2025
- SOC2 claims unverified — No public attestation reports linked
Key Stats
| Metric | Value |
|---|---|
| Pricing | Free (ad/data-supported) |
| Model Support | GPT-5.2, Claude 4.5, Gemini 3 |
| Daily PRs | Thousands |
Cursor
AI-native code editor trusted by NVIDIA (40K engineers), Salesforce (20K developers), and half the Fortune 500.
| Quick Reference | |
|---|---|
| Website | cursor.com |
| Founded | 2022 |
| Funding | $900M+ ($29B valuation) |
Overview
Cursor is the dominant AI IDE — a VS Code fork with integrated AI capabilities from tab completion to fully autonomous agents. April 2026: Cursor shipped its biggest release in years — Cursor 3.0 (Apr 2) with a completely redesigned interface, Composer 2 (Mar 19) with their own multi-agent model, Self-hosted Cloud Agents (Mar 25), and BugBot for automated PR review. The company previously acquired Graphite for code review. With NVIDIA and Salesforce deployments, Cursor is eating the IDE market and expanding up the stack.
Strengths
- Massive adoption — Fortune 500 customers including NVIDIA (40K) and Salesforce (20K)
- Familiar UX — VS Code fork means zero learning curve
- Full autonomy spectrum — Tab completion to full agents in one tool
- Composer 2 — Multi-agent model with self-hosted cloud agent option
- BugBot — Automated PR review expanding beyond code generation
- Model flexibility — OpenAI, Anthropic, Gemini, xAI, and Cursor models
Cautions
- IDE lock-in — Must use Cursor's editor, not just a plugin
- Price escalation risk — Usage-based pricing can surprise with overages
- Individual-focused — Core value is developer speed, not workflow orchestration
Key Stats
| Metric | Value |
|---|---|
| Pricing | Free Hobby, $20/mo Pro, $60/mo Pro+, $200/mo Ultra, $40/user Teams |
| Model Support | 5+ providers |
| Enterprise Customers | Half the Fortune 500 |
Devin
First autonomous AI software engineer with self-scheduling and multi-agent orchestration, trusted by Goldman Sachs, Santander, and Nubank.
| Quick Reference | |
|---|---|
| Website | cognition.ai |
| Founded | 2023 |
| Funding | $175M+ |
Overview
Devin pioneered the "AI software engineer" category in March 2024, setting new benchmarks for autonomous coding. Now with hundreds of thousands of merged PRs, the platform excels at junior-level tasks at scale: migrations, security fixes, test generation. Custom Cognition models (SWE-1.5, SWE-grep) optimize for coding agent workloads.
March 2026 Updates: Cognition shipped two significant capabilities: "Devin can now Schedule Devins" (Mar 20) — enabling scheduled task execution, and "Devin can now Manage Devins" (Mar 19) — multi-agent orchestration where one Devin spins up and manages other Devins. Additionally, Cognition launched "Cognition for Government" (Feb 25) as a new vertical. These moves signal a shift from "AI developer" to "AI engineering manager" — autonomously scheduling and orchestrating sub-agents.
Strengths
- First mover — Established category and brand recognition
- Enterprise traction — Goldman Sachs, Santander, Nubank, Infosys
- Custom models — SWE-1.5 and SWE-grep optimized for coding
- Proven scale — Hundreds of thousands of merged PRs
- Self-scheduling — Devin can now schedule recurring tasks autonomously
- Multi-agent orchestration — Devin managing other Devins for complex workflows
- Government vertical — Cognition for Government expands addressable market
Cautions
- Locked to Cognition — No BYOK option; can't bring your own model. Teams wanting agent flexibility can't swap in Claude Code or Codex
- Poor customer support — Reported slow, unhelpful responses
- High price point — ~$500/seat/month is expensive for many teams (no public pricing page currently — returns 404)
- Struggles with ambiguity — Needs clear requirements
Key Stats
| Metric | Value |
|---|---|
| Pricing | ~$500/seat/mo (pricing page currently offline) |
| Model Support | Custom Cognition models (proprietary) |
| PR Merge Rate | 67% |
| New Capabilities | Self-scheduling (Mar 20), multi-agent mgmt (Mar 19), Gov vertical (Feb 25) |
Entire
Open-source CLI that captures agent sessions per git commit, backed by $60M seed round.
| Quick Reference | |
|---|---|
| Website | entire.io |
| Founded | ~2025 |
| Funding | $60M seed |
Overview
Entire has emerged from stealth with a clear product: an open-source, MIT-licensed CLI that captures agent sessions and ties them to git commits. This creates a provenance layer — teams can see exactly what AI agent work happened for each commit, enabling audit trails, debugging, and knowledge transfer. The $60M seed round signals strong investor conviction. The product occupies a different niche from orchestration platforms (context capture vs. workflow automation) but could evolve into a competitive position.
Strengths
- Massive seed funding — $60M gives significant runway to build and iterate
- Open source (MIT) — Free to use, inspect, and modify
- Unique niche — Agent session capture per commit is a problem no one else is solving
- Composable — Works alongside existing agents rather than replacing them
Cautions
- Different niche — Context capture, not orchestration — may not compete directly with Tembo
- Early product — Just emerged from stealth; maturity unclear
- Adoption risk — Requires developers to add another CLI to their workflow
Key Stats
| Metric | Value |
|---|---|
| Funding | $60M seed |
| License | MIT (open source) |
| Category | Agent session capture / provenance |
Factory
Enterprise Droids with model flexibility and multi-interface support.
| Quick Reference | |
|---|---|
| Website | factory.ai |
| Founded | 2023 |
| Funding | $50M+ Series B |
Overview
Factory builds "Droids" — autonomous coding agents designed for enterprise teams. Unlike single-IDE tools, Droids work across any IDE, CLI, Slack, Linear, or web interface. The model-agnostic approach (Claude, GPT, Gemini) lets teams switch providers without changing workflows. $50M Series B from Wipro Ventures signals enterprise focus.
Strengths
- Enterprise-ready — SOC2, SSO, SAML, audit trails, on-premise options
- Model flexibility — Not locked to single LLM provider
- Workflow integration — Slack, Linear, IDE, CLI access
- Self-improvement — "Signals" system for recursive agent improvement
Cautions
- Crowded market — Many well-funded competitors
- Foundation model risk — OpenAI/Anthropic could replicate features
- Token costs — Usage-based pricing can escalate for heavy use
Key Stats
| Metric | Value |
|---|---|
| Pricing | Pro $20/mo, Max $200/mo, Enterprise custom |
| Model Support | GPT-5, Claude Sonnet 4, Claude Opus 4.1, Gemini |
| Enterprise Features | Full suite |
| New | London office expansion |
Niteshift
Background coding agent delivering merge-ready PRs with browser-verified screenshots.
| Quick Reference | |
|---|---|
| Website | niteshift.dev |
| Founded | ~2025 |
| Funding | Not disclosed |
Overview
Niteshift takes a verification-first approach to autonomous coding: give it a prompt, bug report, or Lovable/Figma prototype, and it returns a merge-ready PR with browser screenshots proving the changes work. The platform runs full cloud dev environments, supports Claude Opus 4.6 and Codex 5.3, and connects to Linear, Notion, and Figma via remote MCP servers with OAuth. Currently in waitlist/pre-launch with active weekly development.
Strengths
- Verification-first — Browser screenshots and API capture prove code works before review
- Prototype pipeline — Converts Lovable/Figma prototypes to production with real API calls
- CI autofix — Automatically fixes CI failures and responds to review comments
- MCP ecosystem — Remote MCP servers via OAuth for extensible integrations
Cautions
- Pre-launch — Waitlist-only with no public pricing or disclosed funding
- Limited traction — No public GitHub repo or community metrics
- Small team — Founder-stage competing against well-funded players
Key Stats
| Metric | Value |
|---|---|
| Pricing | TBA (waitlist) |
| Model Support | Claude Opus 4.6, Codex 5.3 |
| Status | Pre-launch (waitlist) |
Ona
Formerly Gitpod. Full-stack platform for background agents with scheduled and event-driven automations, kernel-level security, and Fortune 500 enterprise traction.
| Quick Reference | |
|---|---|
| Website | ona.com |
| Founded | ~2020 (as Gitpod, rebranded to Ona) |
| Funding | Not disclosed |
Overview
Gitpod has rebranded to Ona and repositioned as "the platform for background agents." This is a significant strategic shift from cloud development environments to AI agent infrastructure. The platform now features Automations ("Agent fleets at scale. Triggered across your codebase with repeatable workflows that run on PRs, schedules, or webhooks") and Veto (kernel-level security for AI agents). Enterprise clients include BNY, Blackstone, Vanta, Pearson, and Hargreaves Lansdown. SOC 2 certified with Fortune 500 customers. Note: "Start for free" still links to app.gitpod.io, indicating the product transition is ongoing.
Strengths
- Full automation support — Scheduled workflows, PR-triggered, and webhook-driven automations — directly competing with Tembo's top differentiator
- Kernel-level security (Veto) — OS-level agent isolation, not just containers
- Enterprise traction — BNY, Blackstone, Vanta, Pearson, Hargreaves Lansdown
- Brand recognition — Gitpod's existing developer mindshare and enterprise relationships
- SOC 2 certified — Compliance-ready for regulated industries
- Scale — Claims 2M developers on platform
Cautions
- Infrastructure-focused — More focused on VPC deployment and environments than workflow orchestration
- Transition in progress — Product still partially on Gitpod infrastructure
- Black box pricing — No public pricing information
- Enterprise-only — Not accessible to individuals or small teams
- Integration depth unclear — Fewer native developer tool integrations (Linear, Raycast) compared to Tembo
Key Stats
| Metric | Value |
|---|---|
| Pricing | Enterprise (custom), free tier available |
| Compliance | SOC 2 certified |
| Developers | 2M claimed |
| Enterprise Clients | BNY, Blackstone, Vanta, Pearson, Hargreaves Lansdown |
OpenCode
Most popular open-source coding agent with 120K GitHub stars, 5M monthly developers, and 75+ provider support.
| Quick Reference | |
|---|---|
| Website | opencode.ai |
| Founded | Unknown |
| Funding | OSS (community) |
Overview
OpenCode is the clear community favorite — 120K GitHub stars, 650+ contributors, 5 million monthly developers. Built by terminal.shop creators, it provides a provider-agnostic alternative to Claude Code and Codex. Available as desktop app (now in beta), terminal CLI, and IDE extensions with support for 75+ LLM providers. Also offers "Zen" — a curated model service. Notably, Blocks already integrates with OpenCode, validating the orchestration layer market that sits above individual CLI agents.
Strengths
- Massive community — 120K stars, 650+ contributors, 5M monthly developers
- Provider freedom — 75+ model providers supported
- Cross-platform — Desktop app (beta), CLI, IDE extensions
- Zen service — Curated model offering for teams that don't want to manage API keys
- LSP integration — Automatic language intelligence out of box
Cautions
- No cloud orchestration — Local-first means no background agent support
- Requires API keys — Users manage own model access
- No enterprise features — No Jira, signed commits, compliance
Key Stats
| Metric | Value |
|---|---|
| Pricing | Free (OSS) |
| Model Support | 75+ providers |
| GitHub Stars | 120K |
| Monthly Developers | 5M |
Replicas
YC-backed background coding agent with multi-surface triggers (GitHub, Slack, Linear, API).[1]
| Quick Reference | |
|---|---|
| Website | replicas.dev |
| Founded | ~2025 |
| Funding | Y Combinator |
Replicas is a managed background coding agent platform where engineering teams delegate tasks from Linear, Slack, GitHub PR comments, or a web dashboard. Each task runs in a sandboxed VM with a full dev environment (dependencies, databases, services), producing pull requests for human review. Teams report shipping 30%+ of PRs through Replicas. The multi-surface trigger model — any team member can kick off work from wherever they already are — is the strongest version of this pattern in the category.
Strengths:
- Multi-surface triggers (GitHub + Slack + Linear + dashboard + API) — lowest friction delegation
- Sandboxed VMs with full dev environments (Redis, Postgres, etc.)
- YC backing provides credibility and runway
- Custom workflow API for programmatic integration
Cautions:
- No public pricing — cost structure opaque
- Closed source — no transparency into agent quality
- Limited public info — no blog, docs, or changelog
- Competitive market with well-funded alternatives (Devin, Factory)
| Detail | Value |
|---|---|
| Agents | Claude Code, Codex |
| Integrations | GitHub, Slack, Linear, REST API |
| Local Dev | No (cloud VMs) |
Tembo
Agent-agnostic orchestration platform with scheduled and event-triggered automations.
| Quick Reference | |
|---|---|
| Website | tembo.io |
| Founded | 2022 |
| Funding | $20M |
Overview
Tembo is the orchestration layer for AI coding agents — delegating work to Claude Code, Cursor, Codex, Amp, or OpenCode through familiar interfaces like Slack, Linear, GitHub, and Jira. Rather than building another agent, Tembo focuses on routing tasks, managing multi-repo operations, and integrating with team tools.
Strengths
- Agent-agnostic — Not locked to one AI provider; switch agents per task
- Full automation support — Scheduled (cron) and event-triggered workflows with MCP server access
- Work from anywhere — @tembo mentions work in Slack, Linear, GitHub, Jira
- Multi-repo coordination — Single task updates API, client libs, and docs
- Transparent pricing — Public tiers from free to enterprise
Cautions
- Credit-based pricing — Heavy users may find per-task credits add up
- Agent quality varies — Results depend on which agent you route to
- Learning curve — Optimal use requires understanding agent routing
Key Stats
| Metric | Value |
|---|---|
| Pricing | $60-200/mo |
| Agent Support | 5+ (Claude Code, Cursor, Codex, Amp, OpenCode) |
| Automations | Scheduled (hourly/daily/weekly/monthly) + event triggers (GitHub, Sentry, Linear, Slack) |
| Enterprise Features | SSO, SLA, BYOK |
Architecture/Pattern Analysis
Platform Categories
| Category | Vendors | Approach |
|---|---|---|
| AI IDE | Cursor | Full editor with AI integrated at every level |
| Autonomous Agent | Devin, Factory, Amp, cto.new, Ona | Independent agents that plan and execute |
| Orchestration Layer | Tembo, 8090, Conductor | Coordinate multiple agents and workflows |
| Infrastructure | Background Agents, OpenCode | Building blocks for custom agent systems |
Model Flexibility Approaches
| Approach | Vendors | Pros | Cons |
|---|---|---|---|
| Multi-provider | Cursor, Factory, Amp, OpenCode, cto.new | Maximum flexibility | Integration complexity |
| Agent-agnostic | Tembo, 8090, Conductor | Future-proof | Depends on underlying agents |
| Custom models | Devin | Optimized for use case | Vendor lock-in |
| OSS-first | OpenCode, Background Agents | Community + control | Self-managed |
Gap Analysis
| Feature | Cursor | Devin | Factory | OpenCode | Tembo |
|---|---|---|---|---|---|
| Multi-model support | ✅ | — | ✅ | ✅ | ✅ |
| CLI-agnostic (any agent) | — | — | — | — | ✅ |
| Background agents | ✅ | ✅ | ✅ | — | ✅ |
| Scheduled automations (cron) | — | ✅ | ✅ | — | ✅ |
| Event-triggered automations | — | ✅ | ✅ | — | ✅ |
| Multi-repo operations | — | — | — | — | ✅ |
| Jira integration | — | — | ✅ | — | ✅ |
| BYOK / self-hosted | ✅ | — | ✅ | ✅ | ✅ |
| Open source | — | — | — | ✅ | — |
| Free tier | ✅ | — | — | ✅ | ✅ |
Surface & Interface Matrix
| Platform | Cloud | Local Dev | Mac App | Mobile App | CLI/TUI | IDE Extension |
|---|---|---|---|---|---|---|
| Tembo | ✅ | — | — | — | — | — |
| Factory | ✅ | — | — | — | ✅ | ✅ |
| Cursor | — | ✅ | ✅ (is the app) | — | — | N/A (is IDE) |
| OpenCode | — | ✅ | ✅ | — | ✅ | ✅ |
| Devin | ✅ | — | — | — | — | — |
| 8090 | ✅ | — | — | — | — | — |
| Amp | ✅ | ✅ | — | — | ✅ | ✅ |
| Conductor | — | ✅ | ✅ | — | — | — |
| cto.new | ✅ | — | — | — | — | ✅ (MCP) |
| Ona | ✅ | — | — | — | — | — |
| Background Agents | ✅ | ✅ | — | — | ✅ | — |
Key insight: Tembo is uniquely both model-agnostic and CLI-agnostic — it orchestrates any coding agent (Claude Code, Codex, Amp, OpenCode) rather than being locked to a single CLI tool. This means teams can swap underlying agents without changing their workflow integration.
Integrations Matrix
Integration depth determines how well a platform fits into existing team workflows.
| Category | Integration | Cursor | Devin | Factory | Blocks | OpenCode | Tembo |
|---|---|---|---|---|---|---|---|
| Source Control | GitHub | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
| GitLab | ✅ | ⚠️ | ✅ | ✅ | ✅ | ✅ | |
| Bitbucket | ✅ | — | ⚠️ | ✅ | ✅ | ✅ | |
| Project Mgmt | Linear | — | ⚠️ | ✅ | ✅ | — | ✅ |
| Jira | — | — | ✅ | ✅ | — | ✅ | |
| Monitoring | Sentry | — | — | ⚠️ | ✅ | — | ✅ |
| Databases | PostgreSQL | — | — | — | ✅ | — | ✅ |
| Supabase | — | — | — | — | — | ✅ | |
| AWS RDS | — | — | — | — | — | ✅ | |
| Communication | Slack | — | ✅ | ✅ | ✅ | — | ✅ |
| Raycast | — | — | — | — | — | ✅ | |
| Knowledge | Notion | — | — | — | ✅ | — | — |
Legend: ✅ = Native integration | ⚠️ = Limited/beta | — = Not available
Integration leaders:
- Tembo — Widest integration set (11 native integrations across all categories) with unique cloud database coverage (Supabase, AWS RDS)
- Blocks — Rapidly expanding (8+ integrations including GitHub, GitLab, Bitbucket, Slack, Linear, Jira, Sentry, Postgres, Notion) — closest to Tembo in breadth
- Factory — Strong project management (Linear, Jira) and communication (Slack)
- Cursor — Source control only (as IDE, relies on git)
- Devin — Slack + GitHub focus
Key differentiators: Tembo has unique cloud database integrations (Supabase, AWS RDS) and Raycast for productivity. Blocks has Notion for knowledge management. Both Tembo and Blocks now cover the core SCM + PM + monitoring + communication categories, making integration depth less of a clear differentiator and shifting the competitive axis toward automation capabilities and agent flexibility.
Automation Capabilities Deep Dive
Automations — the ability to run agents on a schedule or in response to events — separate true autonomous platforms from interactive assistants.
| Platform | Scheduled (Cron) | Event Triggers | Example Use Cases |
|---|---|---|---|
| Tembo | ✅ Hourly/daily/weekly/monthly | ✅ GitHub, Sentry, Linear, Slack events | Daily security scans, auto-fix Sentry errors, PR description generation |
| Factory | ✅ Configurable schedules | ✅ PR events, issue creation | Automated code review, recurring maintenance |
| Devin | ✅ Self-scheduling (Mar 2026) | ✅ Slack/issue triggers + multi-agent orchestration | Scheduled tasks, Devin-managed sub-Devins, government vertical |
| Ona | ✅ Schedules, webhooks | ✅ PR-triggered workflows | Agent fleets at scale, repeatable workflows across codebase |
| Blocks | — | ✅ GitHub Actions failures, PR events, GitLab MR, Bitbucket push | CI gate integration, event-driven agent runs |
| 8090 | ⚠️ Limited | ⚠️ Via integrations | Requirements sync, architecture updates |
| Cursor | — | — | Interactive only |
| OpenCode | — | — | Interactive only |
| Amp | — | — | Interactive only |
| Conductor | — | — | Local parallel execution only |
| cto.new | — | — | Interactive only |
Key distinction: The automation landscape has shifted significantly since March 2026. Four platforms now offer both scheduled and event-triggered automations (Tembo, Factory, Devin, Ona), and Blocks adds event-triggered support. However, the nature of automation varies: Devin's scheduling is proprietary (Devin scheduling Devins — locked to Cognition models), while Tembo and Factory offer open workflows compatible with any agent. Ona brings enterprise-grade automation with Gitpod's infrastructure heritage but focuses more on environments than workflow orchestration. Tembo's 11 native integrations remain the broadest automation surface area.
Bottom line: No single platform does everything. Cursor dominates IDE-based work, Tembo leads in orchestration/integration and automations, OpenCode wins for OSS flexibility, and Factory/Devin target enterprise autonomous execution.
Strategic Recommendations
By Primary Need
| Primary Need | Recommended | Why | Runner-Up |
|---|---|---|---|
| Automated workflows | Tembo | Only platform with cron + event triggers + 11 integrations | Factory |
| Individual productivity | Cursor | Dominant IDE, Fortune 500 proven | OpenCode |
| CLI/terminal workflow | OpenCode | 75+ providers, Mac/Win/Linux apps | Amp |
| Enterprise autonomous agents | Factory | SOC2, model flexibility, Wipro backing | Devin |
| Multi-repo coordination | Tembo | Only platform with native multi-repo | 8090 |
| Database-aware agents | Tembo | Only platform with PostgreSQL/Supabase/RDS | None |
| Jira/Linear integration | Tembo | Native bidirectional sync | Factory |
| Sentry auto-fix | Tembo | Only platform with native Sentry integration | None |
| Budget-conscious | cto.new | Free frontier models | OpenCode |
| Regulated industries | Ona | Bank/pharma deployments, SOC2/GDPR | Factory |
| Open source | OpenCode | 101K stars, full control | Background Agents |
| Mac-native parallel | Conductor | Native app, git worktree isolation | Cursor |
By Buyer Profile
Enterprise DevOps (100+ engineers): → Tembo for automation + integration depth (Sentry, Linear, Jira, Slack, databases) with agent flexibility. Factory if you need SOC2 out of the box. Ona for heavily regulated environments.
Growth-stage startup (10-50 engineers): → Cursor for AI IDE adoption + Tembo for automated workflows (Sentry error fixes, PR generation, security scans). Evaluate cto.new if budget is tight.
Solo developer / indie hacker: → OpenCode for CLI flexibility with zero cost, or Cursor Free for familiar VS Code. cto.new if you want cloud agents without subscriptions.
Teams with existing Linear/Jira workflows: → Tembo or Blocks — both can pick up tickets and route to agents. Tembo adds scheduled automations; Blocks adds Notion integration.
Teams needing database migrations: → Tembo — PostgreSQL, Supabase, and AWS RDS integrations give agents schema context that no other platform offers.
Market Outlook
Near-Term (2026)
- The market is splitting into two camps: (1) IDE-first tools (Cursor, Amp) and (2) orchestration/background agent platforms (Tembo, Blocks, Ona, Conductor, Devin). Camp 2 is exploding with activity and funding ($60M Entire, $22M Conductor, YC Replicas)
- Automations become table stakes — Platforms without cron/event triggers lose enterprise deals
- Integration depth drives adoption — Teams choose based on Linear/Jira/Sentry support, not model access
- Cursor 3.0 reshapes expectations — Self-hosted cloud agents, multi-agent Composer 2, and BugBot raise the bar for what an IDE should do
- Amp’s CLI-only bet — Abandoning IDE extensions is a contrarian signal that CLI-first may be the future for power users
- At least two platforms in this report will be acquired or shut down
Medium-Term (2027-2028)
- Orchestration emerges as distinct category — separate from IDE and Agent segments
- Enterprise standardizes on 2-3 agent combinations with orchestration glue
- Database integrations become differentiator — Schema-aware agents outperform blind ones
- Multi-repo operations become table stakes; single-repo tools relegated to individual use
- BYOK and self-hosted options become mandatory for regulated industries
Long-Term (2029+)
- Integration-poor platforms commoditize — Model access alone insufficient for team adoption
- Agent orchestration becomes as standard as CI/CD pipelines
- Open-source agents match commercial alternatives in capability
- Background agents run continuously in production, triggered by monitoring and PM tools
- CLI-agnostic orchestrators win — Teams want to swap agents without changing workflows
Bottom Line
Model flexibility is now table stakes — the real differentiation is automation depth, integration breadth, and surface coverage.
Leaders by Dimension
| Dimension | Leader | Why |
|---|---|---|
| Automations | Tembo | Cron + event triggers + 11 integrations + MCP support. Devin and Ona now have scheduling too, but Tembo's is agent-agnostic |
| Integrations | Tembo / Blocks | Tembo: 11 native (unique: Supabase, AWS RDS, Raycast). Blocks: 8+ (unique: Notion). Gap is closing fast |
| CLI/Agent flexibility | Tembo / Blocks | Both orchestrate multiple CLI agents. Blocks: 7+ agents. Tembo: 5+ agents |
| AI IDE | Cursor | Half the Fortune 500, $29B valuation |
| Open Source | OpenCode | 101K stars, 75+ providers |
| Enterprise Agents | Factory | SOC2, Wipro backing, self-improving Droids |
| Enterprise Security | Ona | Kernel-level Veto security, SOC 2, Fortune 500 bank/finance deployments |
| Database-aware | Tembo | Only platform with Supabase + AWS RDS (Blocks has Postgres only) |
The Strategic Question
Do you need an agent or an orchestrator?
- Agent (Cursor, Devin, Factory): Best for individual productivity and well-defined tasks
- Orchestrator (Tembo, Blocks, 8090): Best for team workflows, automated pipelines, and multi-tool environments
The platforms that win will be those with the deepest integrations into existing team workflows — not just the best AI model access. Model providers will commoditize; workflow integration won't.
If you're choosing today: Start with your integration requirements (Linear? Jira? Sentry? Slack?), then your automation needs (cron? event triggers?), then your surface preferences (CLI? IDE? web?). Model support is the last filter, not the first.
Research by Ry Walker Research • methodology
Disclosure: Ry Walker is CEO of Tembo, which competes in this category. This report aims for objectivity but readers should note the affiliation.
Sources