1~
2AI agenti se presunuli z vyzkumnych dem do produnkcnich systemu. Ocekava se, ze vice nez 60 % podnikovych AI aplikaci bude do roku 2026 obsahovat agentni komponenty. Ale budovani agentu od nuly  -  sprava smycek nastroju, stavu, pameti, zpracovani chyb a koordinace vice agentu  -  je slozite. Prave proto existuji frameworky.
3~
4V roce 2026 dominuji ctyri frameworky: **LangGraph**, **CrewAI**, **OpenAI Agents SDK** a **Claude Agent SDK**. Kazdy z nich pristupuje k temuze problemu zasadne odlisnym zpusobem: dat modelum LLM schopnost uvazovat, planovat, pouzivat nastroje a spolupracovat.
5~
6## V kostce
7~
8| Aspekt | LangGraph | CrewAI | OpenAI Agents SDK | Claude Agent SDK |
9|--------|-----------|--------|-------------------|-----------------|
10| **Tvurce** | LangChain | CrewAI Inc. | OpenAI | Anthropic |
11| **Architektura** | Zalozena na grafech | Zalozena na rolich | Zalozena na predavani | Autonomni smycka |
12| **Filozofie** | Maximalni kontrola | Tymova spoluprace | Minimalni abstrakce | Dejte agentovi pocitac |
13| **Jazyky** | Python, TypeScript | Python | Python | Python, TypeScript |
14| **Podpora modelu** | Libovolne (OpenAI, Claude, lokalni) | Libovolne | Libovolne (navzdory nazvu) | Pouze Claude |
15| **Hvezdicky na GitHubu** | ~29k | ~40k | ~21k | ~6k |
16| **Nejlepsi pro** | Slozite stavove workflow | Specializaci vice agentu | Routing a triaz | Kodovani a ulohy s praci se soubory |
17~
18## LangGraph: Stavitel grafu
19~
20LangGraph modeluje workflow agentu jako **orientovane cyklicke grafy**. Definujete uzly (funkce, ktere provadeji praci) a hrany (prechody mezi nimi, volitelne podminene). Stav proteka grafem a je uchovavan prostrednictvim checkpointingu.
21~
22Toto je nejexplicitnejsi a nejlepe kontrolovatelny framework  -  kazdy krok zapojujete sami.
23~
24```mermaid
25graph LR
26    Start --> Router[Router Node]
27    Router -->|needs research| Research[Research Node]
28    Router -->|needs code| Code[Code Node]
29    Research --> Synthesize[Synthesize Node]
30    Code --> Synthesize
31    Synthesize --> End
32```
33~
34### Klicove koncepty
35~
36- **StateGraph**: definice grafu s typovanym stavem
37- **Nodes**: funkce Pythonu transformujici stav
38- **Edges**: spojeni mezi uzly, mohou byt podminena
39- **Checkpointing**: vestavena perzistence pro dlouho bezici workflow
40~
41### Priklad kodu
42~
43```python
44from langgraph.graph import StateGraph, MessagesState, START, END
45from langchain_openai import ChatOpenAI
46~
47llm = ChatOpenAI(model="gpt-4o")
48~
49def call_agent(state: MessagesState):
50    response = llm.invoke(state["messages"])
51    return {"messages": [response]}
52~
53def should_continue(state: MessagesState):
54    last = state["messages"][-1]
55    if last.tool_calls:
56        return "tools"
57    return END
58~
59def call_tools(state: MessagesState):
60    # Execute tool calls and return results
61    results = []
62    for tool_call in state["messages"][-1].tool_calls:
63        result = execute_tool(tool_call)
64        results.append(result)
65    return {"messages": results}
66~
67graph = StateGraph(MessagesState)
68graph.add_node("agent", call_agent)
69graph.add_node("tools", call_tools)
70graph.add_edge(START, "agent")
71graph.add_conditional_edges("agent", should_continue, {"tools": "tools", END: END})
72graph.add_edge("tools", "agent")
73~
74app = graph.compile()
75result = app.invoke({"messages": [{"role": "user", "content": "What's the weather?"}]})
76```
77~
78### Silne stranky
79~
80- Jemnozrnna kontrola nad kazdym krokem a prechodem
81- Vestaveny checkpointing a human-in-the-loop
82- Plna parita s TypeScriptem
83- Funguje s libovolnym poskytovatelem LLM
84- Nejlepsi pro slozite workflow s podminenym vetvenim a smyckami
85~
86### Slabe stranky
87~
88- Strma krivka uceni  -  musite rozumet konceptum teorie grafu
89- Upovdany pro jednoduche pripady  -  zakladni agent vyzaduje vice boilerplate kodu nez ostatni frameworky
90- Ladeni toky grafu muze byt narocne bez LangSmith
91~
92### Cenik
93~
94Open-source (MIT). LangSmith (spravovana platforma pozorovatelnosti) ma placene urovne pro produkcni monitoring.
95~
96## CrewAI: Sestavovatel tymu
97~
98CrewAI pouziva lidskou metaforu: sestavite **posadku** specializovanych agentu, z nichz kazdy ma **roli**, **cil** a **pribeh**. Agenti spolupracuji na **ulohach** pomoci **nastroju**, koordinovani **procesem** (sekvencni, hierarchicky nebo konsensualni).
99~
100Predstavte si to jako najimani tymu, kde kazdy clen ma konkretni pracovni pozici a specializaci.
101~
102```mermaid
103graph TD
104    Crew[Crew Manager] --> R[Researcher\nRole: Find data\nTools: WebSearch]
105    Crew --> W[Writer\nRole: Write content\nTools: FileWrite]
106    Crew --> E[Editor\nRole: Review quality\nTools: FileRead]
107    R --> Task1[Research Task]
108    W --> Task2[Writing Task]
109    E --> Task3[Review Task]
110    Task1 --> Task2 --> Task3
111```
112~
113### Klicove koncepty
114~
115- **Agent**: persona s roli, cilem, pribehem a nastroji
116- **Task**: uloha s popisem, ocekavanym vystupem a prirazenym agentem
117- **Crew**: skupina agentu pracujicich spolecne
118- **Process**: strategie provadeni (sekvencni, hierarchicka, konsensualni)
119- **Flow**: vrstva orchestrace rizena udalostmi pro propojeni vice posadek
120~
121### Priklad kodu
122~
123```python
124from crewai import Agent, Task, Crew, Process
125~
126researcher = Agent(
127    role="Senior Research Analyst",
128    goal="Find comprehensive data about the given topic",
129    backstory="You have 10 years of experience in technology research. "
130              "You are thorough and always verify facts from multiple sources.",
131    tools=[web_search_tool],
132    verbose=True,
133)
134~
135writer = Agent(
136    role="Technical Writer",
137    goal="Create clear, engaging technical content",
138    backstory="You write for a developer audience. "
139              "Your articles are practical and include code examples.",
140    tools=[file_tool],
141    verbose=True,
142)
143~
144research_task = Task(
145    description="Research the latest developments in WebAssembly in 2026. "
146                "Focus on WASI, Component Model, and production use cases.",
147    expected_output="A structured research document with key findings and sources.",
148    agent=researcher,
149)
150~
151writing_task = Task(
152    description="Write a blog post based on the research. "
153                "Include code examples and Mermaid diagrams.",
154    expected_output="A complete blog post in Markdown format.",
155    agent=writer,
156    context=[research_task],  # Writer receives researcher's output
157)
158~
159crew = Crew(
160    agents=[researcher, writer],
161    tasks=[research_task, writing_task],
162    process=Process.sequential,
163    verbose=True,
164)
165~
166result = crew.kickoff()
167print(result.raw)
168```
169~
170### Silne stranky
171~
172- Intuitivni abstrakce zalozena na rolich  -  snadne uvazovani
173- Vice nez 100 vestavenych integraci nastroju
174- Sdilena pamet mezi agenty (kratkodoba, dlouhodoba, entitni)
175- Nejvetsi komunita (~40k hvezdicek na GitHubu)
176- Hierarchicky proces s agentem "manazerem", ktery deleguje a validuje
177~
178### Slabe stranky
179~
180- Mensi jemnozrnna kontrola nez LangGraph  -  definujete role, ne presne cesty provadeni
181- Hierarchicky proces muze byt nepredvidatelny, kdyz se agenti neshodnou
182- Ladeni konverzaci vice agentu je tezsi nez u toky jednoho agenta
183~
184### Cenik
185~
186Open-source jadro (zdarma). CrewAI Platform: 99 $/mesic (Teams) az 120 tis. $/rok (Enterprise). Ceny zalozeny na aktivnich posadkach a mesicnich provedenich.
187~
188## OpenAI Agents SDK: Router
189~
190OpenAI Agents SDK (duchovy nastupce Swarm) se zamruje na **predavani**  -  agenti predavaji konverzace jinym specializovanym agentum. Je to nejminimalnejsi framework: agenti, nastroje, predavani a guardrails. To je vse.
191~
192```mermaid
193graph LR
194    User --> Triage[Triage Agent]
195    Triage -->|billing question| Billing[Billing Agent]
196    Triage -->|refund request| Refund[Refund Agent]
197    Triage -->|technical issue| Support[Support Agent]
198    Billing --> Response[Response]
199    Refund --> Response
200    Support --> Response
201```
202~
203### Klicove koncepty
204~
205- **Agent**: model + instrukce + nastroje + predavani
206- **Handoff**: predani jinemu agentovi (modelovano jako nastroj, ktery LLM muze zavolat)
207- **Guardrail**: validace vstupu/vystupu bezici paralelne s agentem
208- **Runner**: provadi smycku agenta
209- **Tracing**: vestavena pozorovatelnost pro vsechna volani LLM, vyuziti nastroju a predani
210~
211### Priklad kodu
212~
213```python
214from agents import Agent, Runner, handoff, InputGuardrail, GuardrailFunctionOutput
215from pydantic import BaseModel
216~
217class SafetyCheck(BaseModel):
218    is_safe: bool
219    reason: str
220~
221async def content_safety(ctx, agent, input_text):
222    result = await Runner.run(
223        Agent(name="Safety", instructions="Check if input is safe. No PII."),
224        input_text,
225        context=ctx,
226    )
227    output = SafetyCheck.model_validate_json(result.final_output)
228    return GuardrailFunctionOutput(
229        output_info=output, tripwire_triggered=not output.is_safe
230    )
231~
232billing_agent = Agent(
233    name="Billing Agent",
234    instructions="You handle billing inquiries. Be precise with numbers.",
235    tools=[lookup_invoice, process_payment],
236)
237~
238refund_agent = Agent(
239    name="Refund Agent",
240    instructions="You process refund requests. Always verify the order first.",
241    tools=[lookup_order, issue_refund],
242)
243~
244triage_agent = Agent(
245    name="Triage Agent",
246    instructions="Route the customer to the right specialist. "
247                 "Ask clarifying questions if needed.",
248    handoffs=[billing_agent, refund_agent],
249    input_guardrails=[InputGuardrail(guardrail_function=content_safety)],
250)
251~
252result = await Runner.run(triage_agent, "I need a refund for order #4521")
253print(result.final_output)
254# The triage agent routes to refund_agent, which processes the refund
255```
256~
257### Silne stranky
258~
259- Cisty vzor predavani  -  prirozeny pro workflow routingu/triaze
260- Guardrails bezi paralelne s provadenim (fail-fast, neblokovaci)
261- Vestaveny panel tracing pro ladeni
262- Navzdory nazvu podporuje modely mimo OpenAI
263- Minimalni abstrakce  -  snadne pochopeni a rozsireni
264~
265### Slabe stranky
266~
267- Mene vyspela sprava stavu nez LangGraph
268- Zadna vestavena perzistence ani checkpointing
269- Ekosystem nastroju tretich stran je mensi
270- Design zamereny na predavani nemusi sednou na kazdou architekturu
271~
272### Cenik
273~
274Open-source (MIT). Platite za token za jakykoli model, ktery pouzivate.
275~
276## Claude Agent SDK: Vyvojar
277~
278Claude Agent SDK zaujima odlisny pristup: namisto definovani workflow nebo roli date agentovi **sadu nastroju a nechate ho vymyslet, jak ulohu splnit**. Pouziva stejnou autonomni smycku, ktera pohani Claude Code  -  cist, jednat, overit, iterovat.
279~
280```mermaid
281graph TD
282    Prompt[User Prompt] --> Loop[Autonomous Agent Loop]
283    Loop --> Reason[Reason about next step]
284    Reason --> Act[Execute tool]
285    Act --> Verify[Check result]
286    Verify -->|not done| Loop
287    Verify -->|done| Output[Final output]
288```
289~
290### Klicove koncepty
291~
292- **query()**: hlavni vstupni bod, ktery spusti smycku agenta
293- **Vestavene nastroje**: Read, Write, Edit, Bash, Glob, Grep, WebSearch, WebFetch
294- **Vlastni nastroje pres MCP**: definice nastroju jako in-process MCP serveru
295- **Sub-agents**: specializovani agenti, kterym muze rodicovsky agent delegovat
296- **Sessions**: udrzovani kontextu napric vice interakcemi
297~
298### Priklad kodu
299~
300```typescript
301import { tool, createSdkMcpServer, query } from "@anthropic-ai/claude-agent-sdk";
302import { z } from "zod";
303~
304const searchDocs = tool(
305  "search_docs",
306  "Search the internal documentation for relevant information",
307  { query: z.string().describe("Search query") },
308  async ({ query }) => {
309    const results = await vectorStore.similaritySearch(query, 5);
310    return {
311      content: [{ type: "text", text: results.map(r => r.pageContent).join("\n\n") }],
312    };
313  }
314);
315~
316const docsServer = createSdkMcpServer({
317  name: "docs",
318  version: "1.0.0",
319  tools: [searchDocs],
320});
321~
322for await (const message of query({
323  prompt: "Find how authentication works in our system and write a summary",
324  options: {
325    mcpServers: { docs: docsServer },
326    allowedTools: ["Read", "Glob", "Grep", "mcp__docs__search_docs"],
327  },
328})) {
329  if (message.type === "result" && message.subtype === "success") {
330    console.log(message.result);
331  }
332}
333```
334~
335### Silne stranky
336~
337- Prvotrodni integrace MCP  -  pripojte se k libovolnemu ekosystemu MCP serveru
338- Vestavene nastroje pro operace se soubory, terminal a pristup na web
339- Automaticka kompaktace kontextu pro velke kodove baze
340- Paralelni sub-agents pro slozite ulohy
341- Stejny motor jako Claude Code  -  overeny v realnych vyvojarskych workflow
342~
343### Slabe stranky
344~
345- Pouze modely Claude  -  zadna podpora vice poskytovatelu
346- Novejsi framework s mensi komunitou
347- Vyzaduje prostredi Node.js i pro Python SDK
348- Mensi explicitni kontrola workflow ve srovnani s LangGraph
349~
350### Cenik
351~
352Open-source. Standardni sazby za tokeny Claude API. Managed Agents (hostovana verze): 0,08 $ za hodinu relace plus naklady na tokeny.
353~
354## Kdy zvolit ktery
355~
356```mermaid
357graph TD
358    Start{What's your priority?}
359    Start -->|Full control over workflow| LG[LangGraph]
360    Start -->|Multi-agent collaboration| CA[CrewAI]
361    Start -->|Routing and triage| OA[OpenAI Agents SDK]
362    Start -->|Coding and file automation| CS[Claude Agent SDK]
363~
364    LG --> LGU[Complex stateful workflows\nConditional branching\nHuman-in-the-loop]
365    CA --> CAU[Team of specialized agents\nResearch + writing pipelines\nContent generation]
366    OA --> OAU[Customer service routing\nMulti-step handoffs\nInput validation]
367    CS --> CSU[Code generation and review\nFile-heavy automation\nMCP tool ecosystem]
368```
369~
370### Zvolte LangGraph, pokud:
371- Potrebujete presnou kontrolu nad kazdym krokem workflow
372- Vas pripad uziti zahrnuje slozitou podminenou logiku a smycky
373- Chcete vestavenou perzistenci a checkpointy human-in-the-loop
374- Potrebujete pouzivat vice poskytovatelu LLM ve stejnem workflow
375~
376### Zvolte CrewAI, pokud:
377- Chcete intuitivni abstrakci zalozenou na rolich
378- Vase uloha zahrnuje vice agentu s odlisnymi specializacemi
379- Potrebujete, aby agenti spolupracovali a predavali si kontext
380- Cenite si nejvetsi komunity a nejvice vestavenych integraci
381~
382### Zvolte OpenAI Agents SDK, pokud:
383- Vasim primarnim vzorem je smerovani konverzaci ke specialistum
384- Potrebujete guardrails validujici vstup/vystup paralelne
385- Chcete co nejjednodussi abstrakci s minimalnim boilerplate kodem
386- Vestaveny tracing a pozorovatelnost jsou dulezite
387~
388### Zvolte Claude Agent SDK, pokud:
389- Vasi agenti potrebuji cist, psat a spoustet kod
390- Chcete prvotrodni integraci s MCP servery
391- Potrebujete autonomni agenty, kteri iteruji a sami se opravuji
392- Jiz pouzivate Claude a chcete nejhlubsi integraci
393~
394## Lze frameworky kombinovat?
395~
396Ano. Castym vzorem je pouziti jednoho frameworku pro orchestraci a druheho pro jednotlive agenty:
397~
398- **LangGraph** pro celkovy graf workflow
399- **CrewAI** pro konkretni uzel vyzadujici spolupraci vice agentu
400- **Claude Agent SDK** pro podzadania souvisejici s kodovanim pres MCP
401- **OpenAI Agents SDK** pro triaz a routing smerem ke klientum
402~
403Frameworky se navzajem nevylucuji. Pouzijte to, co se hodi pro kazdou cast vaseho systemu.
404~
405## Zaver
406~
407Kazdy framework sazi na neco jineho:
408~
409- **LangGraph** optimalizuje pro kontrolu  -  vy rozhodujete o kazdem prechodu
410- **CrewAI** optimalizuje pro spolupraci  -  agenti pracuji jako tym
411- **OpenAI Agents SDK** optimalizuje pro jednoduchost  -  minimalni abstrakce, cista predavani
412- **Claude Agent SDK** optimalizuje pro autonomii  -  dejte mu nastroje a nechte ho pracovat
413~
414Spravna volba zavisi na vasem workflow, vasem tymu a vasem stavajicim stacku. Vyberte ten, ktery odpovida vasemu primarnimu pripadu uziti, dobre se ho naucte a sahnete po dalsich, az narazite na jejich silnou stranku.
415~