v2.34.0GPT-6 Astra is the default — and qwq_reason works again on Qwen3.8-27BWhat changed →
Open-source multi-model orchestration for MCP

Reasoning in steps, not guesses.

TachiBot runs your prompt through structured reasoning — council, jury, planner, step-by-step — so models debate, verify, and reason deeper before they answer. One answer you can trust.

$npx -y -p tachibot-mcp tachibot init
12
Providers
67
Tools
88
Prompt techniques

Stop trusting. Start verifying.

Same question. Different approach. Better answer.

"What breaking changes are in React 19?"
One model
  • ×Generic advice, might miss recent updates
  • ×No sources or documentation links
  • ×Could confuse React 18 vs 19 features
  • ×No way to verify accuracy
Unverified, possibly outdated
TachiBot
1. Run in parallel
OpenAIGooglePerplexityOpenRouter
2. Cross-verify
Models challenge each other, 3–200 rounds
3. Fact-check
Perplexity + Grok, live sources with recency filters
Verified, with official sources
Multi-model deliberation

The Council & Jury

One question fans out to a panel of independent models in parallel. Their perspectives are aggregated, then a Gemini judge synthesizes a single verdict — the “replace judges with juries” pattern, made real.

QUERY
grokfirst-principles
deepseekfrontier math
kimistep-by-step
openaitradeoffs
gemini_judge
synthesis
VERDICT
Judgegemini_judge
One model · one pass

A single evaluator. It scores, ranks, resolves conflicts, or merges perspectives into one answer — the synthesis primitive the jury and council both lean on.

Juryjury
Many models · one round

Many jurors answer in parallel, then a judge synthesizes. One tool call, fast — the default for any “which is right / which is best?” question.

Council/council
Many models · many rounds

Research → adversarial debate → sequential reasoning → synthesis, iterated over rounds. For architecture calls where one round isn't enough.

Who sits on the panel

Where each model is strongest.

Lab-diverse on purpose — different training data, different blind spots. Pick jurors per question; Gemini presides.

Grok 4.5juror

Opus-class flagship — real-time info, configurable reasoning effort, lowest hallucination rate.

DeepSeek V4 Projuror

Open-weight frontier reasoning — top AIME & CodeElo math/CP scores.

Kimi K3juror

2.8T open-weight MoE — the largest open model shipped. Long-horizon agentic coding, 1M context, natively multimodal.

GPT-5.6 soljuror

Agentic workflows & rock-solid tool-use reliability. 54% more token-efficient than 5.5.

Gemini 3.1 Projudge

Long-context multimodal reasoning & science — the presiding judge.

GLM-5.2

Long-horizon coding — SWE-Bench Pro 62.1, Terminal-Bench 81.0; agentic planning juror.

StepFun 3.7 Flash

Near-frontier AIME/SWE-Verified reasoning at flash-tier cost.

ERNIE 4.5 VL

Broad world knowledge & human-preference/arena strength.

Gemini 3.6 Flash

High-speed frontier agentic coding & grounded search, 1M context.

Qwen3.8 Max

Heavy mathematical & formal reasoning at configurable effort, 1M context, multimodal.

MiniMax M3

1M-token context + long-horizon agents, cost-efficient.

Local models

Free offline juror via Ollama / LM Studio — zero token cost, fully private. Runs any local model, incl. Nous Hermes builds. (The Hermes agent itself is model-agnostic — 300+ backends from GPT and Claude to self-hosted.)

researched via grok_search + perplexity_ask
Building blocks for thinking

Mix and match your reasoning pipeline.

Multi-model plannerNew

Bulletproof implementation plans from a council of six models — ground truth, code analysis, decomposition, critique, synthesis, and a judged final plan with verification checkpoints.

planner_maker + planner_runner
  1. 01GrokSearches for ground truth
  2. 02QwenAnalyzes code
  3. 03Kimi K3Decomposes into ordered subtasks
  4. 04GPT-5.6Critiques gaps
  5. 05Qwen3.8 MaxDrafts the synthesis
  6. 06GeminiJudges with quality scores
Multi-model jury

A lab-diverse panel of jurors in parallel — 12 to choose from, including a free offline local juror. Gemini synthesizes the verdict.

jury + gemini_judge
88 prompt techniques

Research-backed patterns — reflexion, SCoT, ReAct, pre_mortem and more. Embedded in tool prompts; preview before executing.

preview → execute
YAML workflows

Chain unlimited steps with variables, dependencies, auto-distillation and an optional AI judge. Debates, research pipelines, code reviews.

workflow + planner
The arsenal

Every tool, filterable.

Filter the full toolset by capability or provider — or search by name. Every tool follows one design rule: a single required string param, all enums optional with defaults.

Capability
Provider
Showing 67 of 67 tools
grok_reasonGrok (xAI)

Deep first-principles reasoning with optional heavy mode.

Reasoninggrok-4.5
openai_reasonOpenAI

Mathematical & high-effort reasoning with GPT-6 Astra at xhigh reasoning effort.

Reasoninggpt-6-astra
qwen_reasonQwen

Heavy mathematical reasoning (Qwen3.8 Max, 1M context).

Reasoningqwen3.8-max
qwq_reasonQwen

Multi-perspective deliberation: optimist, pessimist, expert, contrarian.

Reasoningqwen3.8-max
kimi_thinkingKimi

Multimodal reasoning with a 100-subagent swarm.

Reasoningkimi-k3
perplexity_reasonPerplexity

Reasoning grounded in live web search.

Reasoningsonar
deepseek_reasonDeepSeek

Open-weight frontier reasoning — top AIME/GPQA math & logic chains.

Reasoningdeepseek-v4-pro
glm_reasonGLM (Z.ai)

Agentic reasoning & tool-use planning (SWE-Bench Pro leader).

Reasoningglm-5.2
stepfun_reasonStepFun

Efficient deep reasoning — high AIME/SWE-Verified at flash-tier cost.

Reasoningstep-3.7-flash
ernie_reasonERNIE (Baidu)

Broad-knowledge reasoning with human-preference/arena strength.

Reasoningernie-4.5-vl
local_queryLocal

Query a local model via Ollama/LM Studio/llama.cpp/vLLM — free, offline, private.

Reasoningollama · lm studio
nextThoughtTachiBot

Sequential thinking chain with optional per-step model execution.

Reasoningorchestration
thinkTachiBot

Lightweight internal reasoning log.

Reasoningorchestration
qwen_coderQwen

Agentic code generation. SWE-bench >70%.

Codingqwen3.8-max
qwen_algoQwen

Algorithm analysis: complexity & optimization tiers.

Codingqwen3.8-max
deepseek_algoDeepSeek

Strongest algorithmic review: correctness, Big-O, edge cases, data structures.

Codingdeepseek-v4-pro
qwen_competitiveQwen

Competitive-programming problem solving.

Codingqwen3.8-max
kimi_codeKimi

SWE-focused code generation & bug-fixing leader.

Codingkimi-k3
minimax_codeMiniMax

Single-pass code operations, cost-efficient (1M context).

Codingminimax-m3
grok_codeGrok (xAI)

Code analysis & optimization.

Codinggrok-4.5
grok_debugGrok (xAI)

Specialized debugging assistance.

Codinggrok-4.5
debug_triageGrok (xAI)

Ranked root-cause hypotheses with likelihoods, a discriminating check per hypothesis, and the minimal fix for the leader.

Codinggrok-4.5
openai_code_reviewOpenAI

Comprehensive code review with focus areas.

Codinggpt-6-astra
testgenQwen

Generate runnable tests via Qwen3-Coder-Next — edge cases enumerated first.

Codingqwen3.8-max
diff_reviewTachiBot

Multi-model diff-aware review — Kimi K3 + DeepSeek + GPT-6 Astra scoped to changed lines, Gemini-judged verdict.

Codingorchestration
perplexity_askPerplexity

Web search with current information (Sonar).

Researchsonar
grok_searchGrok (xAI)

Live web search with Grok 4.6.

Researchgrok-4.5
grok_search_liteGrok (xAI)

Cheaper live search on grok-4.3 — high-volume lookups and jury fan-outs.

Researchgrok-4.3
openai_searchOpenAI

Real-time web search via GPT-5.6.

Researchgpt-6-astra
gemini_searchGemini

Web search with Google Search grounding.

Researchgemini-3.7-flash
gemini_analyze_codeGemini

Code quality, security & performance analysis.

Analysisgemini-3.1-pro · 3.7-flash
gemini_analyze_textGemini

Rhetorical analysis: bias, fallacies, persuasion.

Analysisgemini-3.1-pro · 3.7-flash
kimi_long_contextKimi

Long-context analysis across a 1M window.

Analysiskimi-k3
openai_explainOpenAI

Clear explanations for complex topics, any level.

Analysisgpt-6-astra
security_reviewDeepSeek

OWASP/CWE security audit via DeepSeek V4 Pro — taint/data-flow analysis, severity per finding, concrete fixes.

Analysisdeepseek-v4-pro
grok_brainstormGrok (xAI)

Contrarian first-principles idea generation.

Brainstormgrok-4.5
openai_brainstormOpenAI

Find the 3rd, 4th & 5th alternative approaches.

Brainstormgpt-6-astra
gemini_brainstormGemini

Convergent synthesis — clusters ideas into themes.

Brainstormgemini-3.1-pro · 3.7-flash
juryTachiBot

Multi-model jury panel → Gemini judge synthesis.

Councilorchestration
gemini_judgeGemini

Multi-perspective evaluation & verdict synthesis.

Councilgemini-3.1-pro · 3.7-flash
planner_makerTachiBot

Council-based plan creation via coordinator pattern.

Planningorchestration
planner_runnerTachiBot

Execute plans against goal-oriented verification gates.

Planningorchestration
list_plansTachiBot

List recently created plans.

Planningorchestration
kimi_decomposeKimi

Structured task decomposition via agent swarm.

Planningkimi-k3
minimax_agentMiniMax

Multi-step task decomposition & execution.

Planningminimax-m3
grok_architectGrok (xAI)

System architecture & design via 4–16 agent swarm.

Planninggrok-4.5
plan_critiqueTachiBot

Adversarial multi-model plan red-team — pre-mortem, hidden assumptions, ranked risks, Gemini-judged verdict.

Planningorchestration
spec_writerOpenAI

Loose request → reviewable spec via GPT-6 Astra: user stories, Given/When/Then, out-of-scope, open questions. Sign off before planning.

Planninggpt-6-astra
blog_writerKimi

Researched long-form in one call — persona interview for the angle, then plan/draft/restructure/line-edit passes (storm + re3).

Planningkimi-k3
workflowTachiBot

Execute YAML-based multi-step workflows.

Workfloworchestration
workflow_startTachiBot

Begin a workflow session with variables.

Workfloworchestration
continue_workflowTachiBot

Resume an interrupted workflow session.

Workfloworchestration
list_workflowsTachiBot

Discover available workflows.

Workfloworchestration
create_workflowTachiBot

Create a new YAML workflow definition.

Workfloworchestration
workflow_statusTachiBot

Check progress of a running workflow session.

Workfloworchestration
validate_workflowTachiBot

Validate workflow YAML: syntax, tool registry, dependency graph.

Workfloworchestration
validate_workflow_fileTachiBot

Validate a workflow file on disk before running.

Workfloworchestration
visualize_workflowTachiBot

Render a workflow's step graph for inspection.

Workfloworchestration
list_prompt_techniquesTachiBot

Discover techniques — ~9 core by default, all=true for the full 31.

Prompt Eng.orchestration
preview_prompt_techniqueTachiBot

Preview an enhanced prompt before executing, or recommend techniques with technique="auto".

Prompt Eng.orchestration
execute_prompt_techniqueTachiBot

Run a query with a chosen technique applied.

Prompt Eng.orchestration
refine_promptOpenAI

Opt-in prompt improver — raw query → goal-first brief + what-changed + open questions. Never executes.

Prompt Eng.gpt-5.6-luna
tachiTachiBot

Smart auto-router: research, solve, verify, create, judge.

Routingorchestration
focusTachiBot

Mode-based multi-model reasoning (deep, debate, analyze).

Routingorchestration
continue_focusTachiBot

Continue a focus session by its session ID.

Routingorchestration
usage_statsTachiBot

View or reset tool usage & cost statistics.

Routingorchestration
doctorTachiBot

Setup diagnostic — detected API keys, visible vs hidden tools (and why), active profile, suggested first step. Zero-cost.

Routingorchestration

12 providers.
One key, or yours.

Gateway — one OpenRouter key
BYOB — your own provider keys
Perplexity — always needs its own key
One MCP surface with automatic fallbacks. Tools register only when their key is present — including free offline local models.
OpenAI1.05M ctx
gpt-6-astra

Agentic reasoning, code review & brainstorming on the GPT-6 Astra flagship. GPT-5.6 tiers stay explicit — sol for explains, terra for code & search, luna for fast calls.

7 dedicated tools
Gemini1M ctx
gemini-3.1-pro · 3.7-flash

Multi-perspective analysis, brainstorming & the presiding jury judge on 3.1 Pro; search runs the agentic 3.7 Flash.

5 dedicated tools
Grok (xAI)500K ctx
grok-4.5

First-principles reasoning, architecture swarms & contrarian brainstorming. Opus-class flagship with configurable reasoning effort.

8 dedicated tools
Kimi1M ctx
kimi-k3

Largest open-weight model shipped — 2.8T MoE, natively multimodal. Task decomposition, long-horizon agentic coding & 1M-context analysis.

5 dedicated tools
Qwen1M ctx
qwen3.8-max

Algorithm analysis & heavy mathematical reasoning with configurable effort; agentic code generation on Qwen3-Coder-Next.

6 dedicated tools
MiniMax1M ctx
minimax-m3

Cost-efficient coding & long-horizon agentic execution. MSA sparse attention, multimodal, 1M context.

2 dedicated tools
Perplexitylive web
sonar

Live web search & research grounded in current sources. Cost-optimized on Sonar.

2 dedicated tools
DeepSeekopen wts
deepseek-v4-pro

Open-weight frontier reasoning — top AIME/CodeElo math & competitive programming. Default juror & strongest algorithmic reviewer.

3 dedicated tools
GLM (Z.ai)1M ctx
glm-5.2

Long-horizon coding flagship from Zhipu — SWE-Bench Pro 62.1, Terminal-Bench 81.0, usable 1M context.

1 dedicated tool
StepFun196B
step-3.7-flash

Efficient deep reasoning — near-frontier AIME/SWE-Verified scores at flash-tier cost.

1 dedicated tool
ERNIE (Baidu)424B MoE
ernie-4.5-vl

Broad world knowledge & human-preference strength. Multimodal 424B/47B-active MoE.

1 dedicated tool
Localoffline
ollama · lm studio

Any OpenAI-compatible local server — Ollama, LM Studio, llama.cpp, vLLM. Zero-cost, private, offline jurors.

1 dedicated tool
>70%
Qwen3-Coder-Next · SWE-bench
62.1
GLM-5.2 · SWE-Bench Pro
2.8T
Kimi K3 · open-weight MoE
−54%
GPT-5.6 sol · tokens vs 5.5
1M
Qwen3.8 Max · context + effort

Get started in minutes.

Add to Claude Code, Claude Desktop, or any MCP client.

Recommended — detects your keys and clients
npx -y -p tachibot-mcp tachibot init

Prints the exact config · never writes or echoes keys

Open source · built in public · actively maintained

Give your agent a council.

New releases ship regularly. If it helps your workflow, a star helps others find it.

Report a bug or request a feature →
npm downloadsLast commit