Agentic Index
Refact.ai vs Tabby (2026)
Both vendors last verified
Both are open source and self hosted, and the split is agent against assistant. That verdict is the Agentic Index coverage score, graded from each vendor's own published materials.
Refact is an autonomous agent, which reports the top open source result on SWE-bench Verified, that plans, edits, runs tests, and iterates, with zero telemetry self hosting and a Pro plan from $10 a month. Tabby is a fast completion and codebase answer server, lighter to run, with a free Community plan for small teams and a Team plan at $19 per seat monthly for up to 50 users.
On the Agentic Index coding agent ranking, neither Refact.ai nor Tabby clears the bar, which asks for all five merge loop capabilities documented in full. Neither documents testing, debugging and optimization in full, nor observability and auditability. 3 of the 96 vendors in the lane clear it. See the coding agent ranking
This comparison is published by Agentic Index, an independent agentic AI vendor research platform. Refact.ai and Tabby are each graded against the same 14 capability Agentic Index taxonomy, from the vendor's own public materials under the Agentic Index verification standard, alongside 1,140 researched vendors. No vendor pays for placement and no vendor has reviewed this page. How this evidence is graded
Choose Refact.ai if
- You want end to end task completion on your own servers, not just suggestions.
- The SWE-bench Verified record, 352 of 500 issues, is reproducible and open.
- BYOK flexibility across providers or local models runs the agent at raw cost.
Choose Tabby if
- Fast completion and codebase answers cover your actual need, and agents are premature.
- You want the simplest possible self hosting, one binary or container on consumer GPUs.
- The free Community plan covers a team of up to 5, with cost limited to hardware.
| Feature | R Refact.ai |
T Tabby |
|---|---|---|
| Action & orchestration | ||
|
Integrations & Tool Calling Ability to connect agents to real systems through native integrations, OAuth-authenticated actions, custom tools, APIs, webhooks, or MCP-compatible tools. |
||
|
Refact.aiIntegrations & Tool Calling Integrations span GitHub and GitLab for source control, PostgreSQL and MySQL databases, Docker containers, shell commands, the Python debugger, browser tools, and MCP servers the agent consumes, with CI/CD pipelines connected. The agent connects to a customer's GitHub, Docker and PostgreSQL to mimic developer behavior and complete tasks autonomously, such as creating a branch, implementing an endpoint, running tests in a container, updating a database and opening a pull request from a single prompt. Support covers 25+ programming languages. Sourceplugins.jetbrains.com Refact listing, refact.ai and github.com/smallcloudairead 2026-08-30 |
||
|
TabbyIntegrations & Tool Calling Pochi connects to MCP servers, local over stdio or remote over HTTP and SSE, with credentials passed as request headers. Servers are set globally or per project, and each tool can be switched on or off. A Use MCP permission governs when the agent calls databases, APIs and other services through them. Through its GitHub Action, Pochi also comments on and commits to the repository. The self hosted Tabby server takes in context through Context Providers and REST APIs and returns suggestions. Sourcedocs.getpochi.com/mcp and /permissionsread 2026-10-25 |
||
|
Workflow Orchestration Ability to sequence, branch, retry, route, and combine deterministic workflow nodes with autonomous agent steps. |
||
|
Refact.aiWorkflow Orchestration The agent plans a task, searches and analyzes the repository for context, edits files, reproduces problems, runs tests and iterates until the work passes, executing up to thirty autonomous actions per task and revisiting earlier steps on failure, with internal sub-agents for planning and for debugging that drive the Python debugger to modify and run scripts; it carries a task from a single prompt through branch creation, implementation, containerized test runs and an opened pull request. Refact reports ranking first among open-source agents on SWE-bench Verified, autonomously resolving 352 of 500 real GitHub issues, with the benchmark pipeline published as open source. Sourcegithub.com/smallcloudai, refact.ai and refact.ai/blog/2025/open-source-sota-on-swe-bench-verified-refact-airead 2026-08-30 |
||
|
TabbyWorkflow Orchestration Pochi, TabbyML's coding agent, runs several agents at once, each in its own Git worktree, and each task opens in its own tab with a diff against main. Since September 2026 a task can also hand independent parts to multiple subagents that work while the main conversation continues. Custom agents can launch other agents as new tasks. Pochi carries a Research Preview label but installs publicly from the VS Code Marketplace and OpenVSX. The self hosted Tabby server itself serves single turn completion, answers and chat. Sourcedocs.getpochi.com/parallel-agents and /developer-updatesread 2026-10-25 |
||
|
Triggers & Channel Coverage How agents wake up and where they work: schedules, webhooks, message events, CRM events, inbox events, chat, email, voice, and collaboration tools. |
||
|
Refact.aiTriggers & Channel Coverage The engine's scheduler runs cron-style scheduled tasks that fire into the agent: durable jobs persist per project in .refact/scheduled_tasks.json and survive restarts, use 5-field cron expressions with jitter, catch up missed one-shot runs and expire recurring jobs after 30 days by default. Invocation otherwise spans VS Code, JetBrains, Visual Studio, Neovim and Sublime Text plugins, a standalone CLI and TUI, and an in-browser GUI served by the local daemon. Sourcegithub.com/JegernOUTT/refact/wiki Scheduler-and-Cron and Home, reached through the docs.refact.ai redirectread 2026-09-29 |
||
|
TabbyTriggers & Channel Coverage Pochi's GitHub Action runs when a comment starts with /pochi, and workflows can post that comment on new pull requests, pushes, new issues and a cron schedule. Pochi then reviews, fixes and commits, and its verdict can gate tests and deploys. The Action needs a Pochi API key from app.getpochi.com, the cloud account TabbyML still offers through a Research Preview waitlist. Inside a running task, Pochi can watch log messages or CI checks and resume when a new event arrives. Developers reach the self hosted Tabby server through plugins for VS Code, JetBrains, Vim, Neovim and Emacs. Sourcedocs.getpochi.com/tutorials/pochi-github-actions and tabbyml.com/agentread 2026-10-25 |
||
| Knowledge & context | ||
|
Knowledge Grounding & RAG Ability to ground agent behavior in company data through document ingestion, retrieval, external knowledge APIs, semantic search, or RAG layers. |
||
|
Refact.aiKnowledge Grounding & RAG A Rust engine runs inside the IDE as a language server and keeps abstract syntax tree and vector database indexes of the codebase current, exposing symbol and semantic search as tools the model calls, with retrieval-augmented generation grounding both agent work and completion in project context. Refact says the agent understands company context and standards by analyzing documentation and codebase, and enterprise deployments can fine-tune models on the organization's own code. Context sources named on the product page span workspace, codebase, databases, files, documentation and web. Sourcegithub.com/smallcloudai and refact.airead 2026-08-30 |
||
|
TabbyKnowledge Grounding & RAG The server keeps a persistent index of the customer's repositories, parsed with Tree-sitter and embedded. Background jobs reindex it, and v0.32.0 extended it to multiple branches. Context Providers, which are data connectors, add documentation, configuration files, Git repositories and external interfaces, and custom documentation can be added over REST APIs. Completions, the Answer Engine and chat all retrieve from that index. Sourcegithub.com/TabbyML/tabby release feed and tabbyml.comread 2026-09-29 |
||
|
Memory & State Persistence Ability to persist context across a run, conversation, workflow, user, team, or longer-term memory layer. |
||
|
Refact.aiMemory & State Persistence Durable project memory lives in the project scoped .refact/ directory: typed task memories (decision, spec, gotcha, risk, handoff, progress, postmortem, brief) preserve structured task knowledge across sessions for project scoped recall, and a knowledge graph of facts and relationships has staleness tracking and a cleanup pass that removes stale or duplicate nodes. The files sit in the project on disk. The VecDB and AST indexes are separate from this memory and ground the agent in the codebase. Sourcegithub.com/JegernOUTT/refact/wiki Memory-and-Knowledge, reached through the docs.refact.ai redirectread 2026-09-29 |
||
|
TabbyMemory & State Persistence Pochi keeps a project memory file, MEMORY.md, that it writes and carries forward across tasks, on by default since May 2026. A separate task memory file survives context compaction within a task. No lifetime or retention is stated for the project memory. In the self hosted Tabby server, Answer Engine threads persist as shareable Pages and repository indexes persist between sessions. Sourcedocs.getpochi.com/developer-updatesread 2026-10-25 |
||
| Control & trust | ||
|
Human Oversight & Guardrails Approval steps, consent checkpoints, escalation rules, structured guardrails, policy constraints, and pause/resume controls. |
||
|
Refact.aiHuman Oversight & Guardrails Tool execution is mediated by confirm rules in each mode's configuration: safe reads default to auto and destructive or sensitive operations to ask, and when a tool requires confirmation execution pauses until the user approves or rejects it; patch, create, update, remove, move and undo style operations require confirmation. Shell commands and integrations carry allow, ask and deny rules, and the agent is capped at thirty autonomous actions per task. Sourcegithub.com/JegernOUTT/refact/wiki Agent-Tools, Integrations-Shell-Commands and Integrationsread 2026-09-29 |
||
|
TabbyHuman Oversight & Guardrails Pochi sorts its actions into Read, Write, Execute, Use MCP and Retry permissions, and each can be auto approved or left to the developer. With approval on, the developer sees the exact change before approving it, and nothing is written to disk until then. All five are auto approved by default. In the self hosted Tabby server, the developer accepts or rejects every completion, chat answer and inline edit. Sourcedocs.getpochi.com/permissions and /developer-updatesread 2026-10-25 |
||
|
Security, Identity & Governance RBAC, SSO, auditability, encryption, least-privilege tool access, compliance posture, and data handling policy. |
||
|
Refact.aiSecurity, Identity & Governance Access can be restricted to particular files or projects so private or confidential code is withheld from the assistant, Refact says it never stores code server side and offers privacy settings, self hosted deployment runs on premise or in a private cloud with zero telemetry, and self hosted deployments keep audit logs for data access. With hosted Refact Cloud retired, there is no vendor side service retaining customer code. Refact does not describe a security attestation or certification, SSO, SAML or role based access control at the organization level. Sourcedocs.refact.ai, refact.ai and refact.ai/blog/2026/refact-cloud-is-shutting-downread 2026-08-30 |
||
|
TabbySecurity, Identity & Governance Team deployments support LDAP directory authentication, shipped in v0.24.0, and generic OAuth, added in v0.32.0. An admin dashboard manages user accounts, issues API tokens per developer and handles model configuration. Because Tabby is self hosted, code and inference never leave the customer's infrastructure and no external service holds data. Tabby names no security attestation or certification, SAML, SCIM, fine grained role based access control or audit logging. Sourcegithub.com/TabbyML/tabby and tabby.tabbyml.com/docsread 2026-08-30 |
||
|
Observability & Auditability Traces, logs, execution histories, metrics, audit events, and debugging detail for production agent behavior. |
||
|
Refact.aiObservability & Auditability Each chat's trajectory, the chronological agent and chat history including tool execution, is persisted locally under .refact/trajectories and can be searched by the agent, reasoning is shown step by step in the editor, and proposed patches are inspectable before they apply; self hosted deployments keep audit logs for data access. These are local session records on the developer's machine, with no audit log of agent actions, telemetry export or run analytics for a team. Sourcegithub.com/JegernOUTT/refact/wiki Chat-System and Memory-and-Knowledge, and refact.airead 2026-09-29 |
||
|
TabbyObservability & Auditability Every Pochi run is saved as a task, stored locally or in the cloud for signed in users, and can be shared by link. Completed steps show how long they took, and browser sessions can be recorded. No admin view or audit log across a team's tasks is described. The self hosted Tabby server has an admin dashboard with user accounts, per developer API tokens and usage analytics. Sourcedocs.getpochi.com/vscode and /developer-updatesread 2026-10-25 |
||
|
Deployment & Data Residency Deployment modes and options, including SaaS, dedicated cloud, VPC, on-prem, hybrid, local runtime, and self-hosting. |
||
|
Refact.aiDeployment & Data Residency The product is open source under a permissive license and local-first by design, running from the IDE with a local Refact engine and whichever providers or local runtimes the customer configures, with no hosted account or managed inference service required; on-premise and private cloud deployment keep code on the customer's own servers with zero telemetry, and local runtimes including Ollama, LM Studio and vLLM are supported. Hosted Refact Cloud is being retired, and Refact says the project continues as open source, local first, BYOK and community maintained, with local configuration and project data remaining under the customer's control and requests sent only to configured providers and integrations. Sourcerefact.ai/blog/2026/refact-cloud-is-shutting-down and refact.airead 2026-08-30 |
||
|
TabbyDeployment & Data Residency The code is written in Rust, and Tabby is distributed under Apache 2.0 as a self contained server that needs no external database management system or cloud service. It installs by Docker, standalone binary, Homebrew tap or Kubernetes, and runs on consumer grade hardware across NVIDIA CUDA, Apple Metal and CPU. Teams without local hardware can deploy it on GPU cloud hosts. Tabby runs on the customer's own terms, in the cloud or on premises, so code and inference stay inside the customer boundary. Sourcegithub.com/TabbyML/tabby and tabbyml.comread 2026-08-30 |
||
| Solution readiness | ||
|
Prebuilt Agents, Templates & Packs Ready-made workflows, packaged employees, templates, blueprints, industry solutions, and role-specific agents that reduce time-to-value. |
||
|
Refact.aiPrebuilt Agents, Templates & Packs An extensions marketplace discovers and installs three kinds of item, skills, commands and subagents, from built-in embedded sources, built-in GitHub sources and user-added GitHub sources, with the source index bundled in the engine. The catalog includes ready subagents a customer can install. Sourcegithub.com/JegernOUTT/refact/wiki Marketplace, reached through the docs.refact.ai redirectread 2026-09-29 |
||
|
TabbyPrebuilt Agents, Templates & Packs Pochi loads skills from project and global folders, including the shared .agents folder that other coding agents also read. Community skills install from the third party skills.sh registry. Developers define custom agents with their own prompts and toolsets. TabbyML publishes no catalog of its own agents, templates or skills, and the self hosted Tabby server ships one assistant. Sourcedocs.getpochi.com/skills and tabbyml.com/agentread 2026-10-25 |
||
| Platform extensibility | ||
|
Model Flexibility & Routing Ability to work across multiple foundation models, route tasks to different models, or let buyers bring their own providers and keys. |
||
|
Refact.aiModel Flexibility & Routing The agent runs on bring-your-own-key or local model providers configured on the customer's own machine, with at least one provider required in Provider Setup before use; named BYOK providers are OpenAI, Anthropic, Gemini, DeepSeek, OpenRouter, Groq and xAI, and named local runtimes are Ollama, LM Studio, vLLM and custom OpenAI-compatible endpoints, with default models selectable per surface. Code completion ships on a bundled Qwen2.5-Coder model. With Refact issued model credits retired, BYOK or a local runtime is the only way to run the agent. Sourceplugins.jetbrains.com Refact listing and refact.ai/blog/2026/refact-cloud-is-shutting-downread 2026-08-30 |
||
|
TabbyModel Flexibility & Routing The customer selects the model at launch through a serve flag, and a separate chat model can be paired with the completion model in the same command. Embeddings are configurable independently. Supported families include StarCoder and StarCoder2, CodeLlama, Qwen and Qwen2.5-Coder variants, and Codestral through a Mistral integration. Mistral Embedding API support arrived in v0.32.0, alongside Llamafile deployment integration, and teams can fine tune on their own code. Because the server is self hosted, the customer supplies and runs the weights directly. Tabby names no routing gateway that sends each task to a different provider. Sourcegithub.com/TabbyML/tabby and tabby.tabbyml.com/docsread 2026-08-30 |
||
|
APIs, SDKs & MCP Extensibility Composability layer: stable APIs, SDKs, MCP tool consumption/serving, custom tools, and integration into internal systems. |
||
|
Refact.aiAPIs, SDKs & MCP Extensibility The local engine exposes a v1 HTTP API covering chat, code, tools, AST, retrieval, git, checkpoints, integrations, knowledge and voice, and the GUI and IDE plugins are thin clients of that daemon, so other programs can drive Refact the same way. The engine and plugins are open source, and MCP servers the customer configures extend the agent. Sourcegithub.com/JegernOUTT/refact/wiki HTTP-API and Home, reached through the docs.refact.ai redirectread 2026-09-29 |
||
|
TabbyAPIs, SDKs & MCP Extensibility The self hosted server publishes an OpenAPI reference for its completion, chat, event and health endpoints, which the official plugins and cloud IDEs consume. Since v0.29, custom documentation and context sources can be added over REST APIs. The reference is thin and labeled for an older release, and no SDK or MCP server is named. Pochi adds API keys for its GitHub Action and a CLI that runs a task from a prompt. The server and plugins are Apache 2.0 open source. Sourcetabby.tabbyml.com/api and github.com/TabbyML/tabbyread 2026-10-25 |
||
|
Testing, Debugging & Optimization Testing, debugging, scoring, retries, fallbacks, quality gates, and optimization loops for improving agent workflows before and after deployment. |
||
|
Refact.aiTesting, Debugging & Optimization The agent reproduces a reported problem, runs the codebase's existing tests to avoid regressions, and iterates when they fail, with a dedicated debugging sub-agent that drives the Python debugger to modify and run scripts and isolate the cause of a fault; a standalone AI Code Review tool analyzes code and proposes fixes. Refact publishes SWE-bench Verified results with the full benchmark pipeline open sourced for reproducibility. Running the project's tests checks the agent's own output, SWE-bench is Refact's own evaluation, and Refact does not offer customers evals, regression testing or a way to debug agent behavior. Sourcerefact.ai/blog/2025/open-source-sota-on-swe-bench-verified-refact-ai and refact.airead 2026-08-30 |
||
|
TabbyTesting, Debugging & Optimization Before a Pochi task with open todos can finish, a read only audit subagent inspects the workspace, reading files, test output and command results. Those checks lean on the project's own tests rather than a harness of Pochi's. A GitHub Action workflow can also gate tests and deploys on Pochi's approve or reject verdict, and the Action needs the waitlisted Pochi cloud account. The self hosted Tabby server ships no test, review or evaluation tools. Sourcedocs.getpochi.com/developer-updatesread 2026-10-25 |
||
| Specialist automation | ||
|
Browser & Computer Use Browser, desktop, or remote/local computer control for workflows that cannot be handled through stable APIs alone. |
||
|
Refact.aiBrowser & Computer Use Refact includes a built-in Chrome tool: the agent launches Chrome or Chromium, connects to an existing DevTools endpoint or reuses a session for the chat, opens and switches tabs, navigates, inspects and interacts with pages, extracts data and collects screenshots, for debugging web apps and checking UI behavior. A customer account on Refact's homepage describes the agent operating a WordPress admin interface to fix a plugin fault. Sourcegithub.com/JegernOUTT/refact/wiki Browser-Automation and refact.airead 2026-09-29 |
||
|
TabbyBrowser & Computer Use Pochi's browser agent navigates websites, extracts content and interacts with pages, in a managed headless browser or in the developer's own running Chrome. Since May 2026 it connects to local Chrome over the Chrome DevTools Protocol and works inside that session, saved logins included. Sessions can be recorded, and the CLI drives the browser too. The self hosted Tabby server itself has no browser or code execution. Sourcedocs.getpochi.com/developer-updatesread 2026-10-25 |
||
Pricing snapshot
Sourced from the Index pricing dataset · open each vendor's profile for full detail.
| Pricing | R Refact.ai |
T Tabby |
|---|---|---|
|
Entry price Lowest public entry point |
Free for personal and hobby projects, with limited daily agent use and unlimited completions. Pro is from $10 a month with 40 agent requests a day, and Enterprise is by contact. Hosted Refact Cloud is being retired with no final date published, and the open source local engine with the customer's own keys is free. | Community is free and open source for up to 5 users with local deployment; Team is $19 a month per seat for up to 50 users with flexible deployment; Enterprise is custom with unlimited users, SSO and group management |
|
Pricing confidence How public the numbers are |
Public, exact | Public, exact |
|
Billing Primary billing axis |
A monthly subscription limited by daily agent requests. The free plan allows limited daily agent use with unlimited completions, and Pro at $10 a month allows 40 agent requests a day. Bringing your own model key runs the open source engine with no Refact charge, so only the model provider bills. | Per seat for Team, custom for Enterprise; the Community plan is free for up to 5 users, and the team still pays for the hardware that serves the models |
|
Variable cost Workload / overage exposure |
High variable cost | Low variable cost |
|
Free tier / trial Try before you buy |
Free tier
|
Free tier
|
|
Buying motion Self-serve vs sales call |
Self-serve | Mixed |
|
Pricing verified When this pricing was last checked |
More comparisons with Refact.ai or Tabby
Other matchups in coding agents
Not the pairing you were after? These compare a different set of coding agents on the same 14 capabilities.