Every source GILD analyzed, newest first. Each says who the source is, how it knows, whether it still applies and which findings it counts toward. Open a count to see only its sources.
Source
69 sources in All practitioners, with product documentation Showing all 69 sources
No opinion stated
Two agents set up at the same time can fail with RuntimeError: dictionary changed size during iterationGitHub issue
A flow class builds and caches its definition without a lock on first use, so two instances created at once can crash.
No opinion stated
A failed task can be skipped after checkpoint restore because its output remains setGitHub issue
When a task's output file write fails, the checkpoint still stores its output, and a restore can skip the failed task as if it completed.
No opinion stated
[chore]: rm crewai integrationGitHub pull request
Browserbase removed Stagehand's CrewAI integration as stale, with its source, docs and logo, and pointed the old CrewAI docs pages to the main integrations page.
No opinion stated
Bedrock prompt_tokens excludes cache reads and writes, unlike other native providersGitHub issue
For a call with 1,200 prompt tokens, 1,000 of them cache reads, the Bedrock provider reported 200 prompt tokens while other providers reported 1,200.
Mixed
Post on once, an idempotency library for agent tool callsPost on X
Built once, a Python library that keys each agent tool call so retries do not repeat paid calls, with support for CrewAI and other frameworks.
No opinion stated
UploadCache sync wrappers deadlock 30s inside a running event loopGitHub issue
Every synchronous upload-cache call made inside a running event loop froze for 30 seconds and then failed with a bare timeout.
Positive
I built an open-source app where LLM agents debate a stock before a manager agent decides the tradeReddit post
Runs analyst, bull, bear, judge, and manager agents locally on FastAPI and CrewAI with any OpenAI-compatible API; a first session on DeepSeek costs about US$0.01.
Positive
Post on building a research and writing crewPost on X
Says CrewAI feels like managing a team; built a pipeline where a research agent searches the web and a content agent writes the blog post.
Mixed
Agentic AI Frameworks Compared 2026: LangGraph Winsdev.to post
Names CrewAI the best start for beginners because role-based crews map onto how people describe work, lists coarser control, moving abstractions and docs churn as the cost, and advises testing run latency early on complex crews.
"A: CrewAI, because role-based crews map onto how people already describe work."
No opinion stated
deps!: drop the crewai extra to remove chromadb from the lockfileGitHub pull request
Headroom dropped its crewai extra because every CrewAI release requires ChromaDB, which carried unpatched advisories into its lockfile and Dependabot alerts. The change removed 42 packages, and users now install CrewAI beside Headroom.
Negative
Ran one build on eight agent platforms. Two finished.Reddit post
CrewAI and LangGraph both reported a Slack integration as connected and delivered nothing, with no error and nothing in the channel.
Mixed
GPT-6 Sol and Luna agents with tools worked through Bedrock, but every GPT-6 Astra agent with a tool failed with a 400, because CrewAI's retry sent a reasoning setting Astra rejects. Setting the Responses API worked.
Mixed
Comment on: LangGraph or CrewAI for agent handoffsReddit comment
CrewAI feels more natural for role-based workflows, but LangGraph's explicit state, checkpoints, and retries make complex handoffs easier once context grows.
Mixed
I built the same agent in Strands, LangGraph, and CrewAI and recorded every LLM callPersonal blog post, dev.to
CrewAI produced byte-identical output across runs and was quick to prototype, but its install weighed 699 MB across 142 packages against 71 MB for LangGraph.
Mixed
LangGraph vs CrewAI vs OpenAI Agents SDK: Picking Your Agent Framework in 2026dev.to post
Finds CrewAI the shortest route to a working multi-agent prototype, citing published comparisons that put one under 20 lines of Python.
Counts toward
No opinion stated
ToolUsage retries a failed tool call twice per outer attempt, not onceGitHub issue
A failed tool call is retried inside each outer attempt as well as by the outer loop, so a failing tool runs more often than the retry setting suggests.
No opinion stated
LanceDBStorage.delete() ignores scope/older_than filters and causes accidental mass deletionGitHub issue
Memory delete ignored scope and age filters in some cases and removed records outside the requested scope.
Counts toward, pending a recheck
Mixed
Comment on: An Accidental BlackboardHacker News comment
Left to work it out among themselves, CrewAI agents collaborate awkwardly; defining coordination in the initial spec helps, though agents still drift from it.
No opinion stated
Memoized CrewBase methods retain per-instance Agent/Task/Crew in global cacheGitHub issue
Creating a crew per request left every Agent, Task, and Crew object in a module-level cache, so process memory grew with request count.
Counts toward, pending a recheck
Mixed
Post on graph engineering in production multi-agent buildsPost on X
Builds production multi-agent systems on CrewAI, LangGraph, and Agno, and says orchestration structure with human oversight is what keeps dozens of agents under control.
No opinion stated
crew.usage_metrics multiplies token usage by the number of agents sharing an LLM instanceGitHub issue
With two agents sharing one LLM instance, crew.usage_metrics reported 310 tokens for a run that used 155, double the real count.
Negative
I delegated a review in CrewAI. The reviewer ran an export tool the orchestrator never had.Personal blog post, dev.to
With allow_delegation=True, an orchestrator holding only a query tool delegated to a coworker that then ran an export tool the orchestrator never had.
No opinion stated
Agent.kickoff_async guardrail retry calls sync invoke inside event loopGitHub issue
When a guardrail rejected an async agent's first output, the retry ran through the synchronous path and passed a coroutine on as the result, crashing the run.
No opinion stated
refactor: delete vendored CrewAi and mem0 librariesGitHub pull request
EpicStaff removed CrewAI as a dependency of its agent platform, deleting its vendored copies of CrewAI and mem0 in a change of 180,251 deleted lines across 1,403 files, merged Sep 4, 2026.
Not counted
EpicStaff builds its own agent-flow platform, an alternative to CrewAI's. Kept in the record; it takes part in no figure.
No opinion stated
Truncated responses are detected in the Bedrock provider onlyGitHub issue
Only the Bedrock provider flagged responses cut off at the token limit; other providers passed truncated output on as if complete.
Mixed
AI Agent Frameworks Compared 2026: LangGraph vs CrewAI vs OpenAI SDKdev.to post
Rates CrewAI the fastest of six frameworks from idea to working prototype, and finds its role-playing abstraction harder to predict and debug in production, because agents make their own delegation decisions.
"Simple agent patterns require significantly more code in LangGraph than in frameworks designed for rapid prototyping."
Negative
CrewAI's quadratic context problem: why a 5-agent crew costs 6x more than you expectPersonal blog post, dev.to
With default context passing, each task in a sequential crew re-sends earlier outputs, so a five-agent crew uses about six times the tokens of isolated tasks; scoping the context parameter fixes it.
Mixed
Why I'm Betting on CrewAI for Multi-Agent Orchestration (And Where It Falls Short)Personal blog post, dev.to
Bets on CrewAI for research, analysis, and content pipelines with explicit roles and tasks, and lists token costs that compound with every agent call as a real limitation.
No opinion stated
Remove crewai to fix CRITICAL chromadb vulnerabilityGitHub pull request
Removed crewai from the project's framework test extras to drop ChromaDB, whose critical pre-authentication code injection advisory GHSA-f4j7-r4q5-qw2c had no patched version. CrewAI's smoke tests skip until ChromaDB ships a fix.
No opinion stated
Tool exceptions are caught and replaced with generic error messagesGitHub issue
CrewAI replaced tool exceptions with a generic error string and dropped the type, message, and traceback, so the cause of a failure could not be found.
Counts toward, pending a recheck
Positive
Comment on: Show HN, RiskKernel, kill -9 an AI agent and resume it without paying twiceHacker News comment
Runs an agentic system built on CrewAI and planned to try a tool because it supports CrewAI.
Positive
Comment on: building agent workflows across coding agentsHacker News comment
Tests agent instructions in OpenCode, then ports the logic into LangGraph or CrewAI on Amazon Bedrock AgentCore, which takes about an afternoon to set up for production and then runs with little intervention.
"…port the agent logic into a framework like LangGraph or CrewAI and run everything on Amazon Bedrock AgentCore."
No opinion stated
input_files (PDFFile) are passed as base64 via read_file tool, causing context overflowGitHub issue
PDF inputs reached the agent as base64 text through the read_file tool where native file inputs were expected, overflowing the context.
Counts toward, pending a recheck
No opinion stated
cache_breakpoint injected into messages for non-Anthropic providersGitHub issue
A cache_breakpoint flag meant for Anthropic was sent to every provider, and Groq and other OpenAI-compatible APIs rejected the requests.
Counts toward, pending a recheck
No opinion stated
Tool re-execution on task retry has no idempotency guardGitHub issue
When a task is retried, every tool that already ran runs again, with no way to detect a completed call, so payments, emails, or trades can repeat.
Mixed
Comment on: How many products does Microsoft have named 'Copilot'?Hacker News comment
Says CrewAI, LangGraph, and similar frameworks solve how agents pass data but not which agent has authority over which domain or what happens when two agents disagree.
Negative
Comment on: Show HN, I tested 11 AI frameworks for basic securityHacker News comment
Found no runtime injection detection, PII masking, or audit trail in 11 frameworks including CrewAI, and wrote a layer that patches them.
Mixed
Comment on: Tell HN, Litellm 1.82.7 and 1.82.8 on PyPI are compromisedHacker News comment
Pointed out that CrewAI imports LiteLLM directly and falls back to it when the OpenAI, Anthropic, and other provider SDKs are not installed.
Negative
Comment on: Tell HN, Litellm 1.82.7 and 1.82.8 on PyPI are compromisedHacker News comment
Noted that CrewAI pinned LiteLLM to 1.82.6, the last good version, with a commit message that did not mention the compromise, and asked why users were not warned.
Mixed
One Chat App, 4 AI Frameworks ComparedCompany engineering blog
CrewAI's version of the app took about 420 lines against about 160 for Pydantic AI, with research and writer roles out of the box, but needed a background thread for streaming and run_in_executor for async.
Negative
Why Checkpoints Aren't Durable Execution: LangGraph, CrewAI, Google ADK, and OthersCompany engineering blog
CrewAI's task replay keeps only the last kickoff and @persist saves flow state to SQLite without resuming on its own, so the team builds recovery itself.
Not counted
Diagrid sells a durable workflow runtime that the post recommends over CrewAI persistence. Kept in the record; it takes part in no figure.
Negative
Comment on: Does anyone use CrewAI or LangChain anymore?Hacker News comment
Answered the question of whether anyone still uses CrewAI or LangChain with a flat no, calling them bad.
Negative
Comment on: Show HN, Intellistant, a C++ alternative to LangChain for AI agentsHacker News comment
Built a C++ framework after finding Python frameworks such as LangChain and CrewAI sluggish in tool execution and agent coordination on local models.
No opinion stated
Token tracking race condition in threading-based async_executionGitHub issue
Per-task token tracking captured counts before tasks ran in parallel threads, so concurrent tasks were credited each other's tokens.
Counts toward, pending a recheck
Mixed
Comment on: Critical vulnerability in LangChain, CVE-2025-68664Hacker News comment
Found none of the frameworks ready and chose Spring AI; CrewAI was the second choice, though the commenter saw lock-in and paid features behind API keys across these frameworks.
Positive
Comment on: I added native time awareness to CrewAI to fix LLM date hallucinationsHacker News comment
Runs agents in production and contributed a TimeAwarenessTool and an IDGenerationTool to the core CrewAI repository to fix date hallucination and support log tracing.
Mixed
Pydantic AI with Samuel ColvinPodcast episode (Software Engineering Daily)
Looked at LangChain, LangGraph, CrewAI and LlamaIndex when Pydantic started building AI features into Logfire, found them short on type safety and engineering standards, and built Pydantic AI instead.
"Obviously, LangChain, LangGraph, Crew AI, LlamaIndex, all of these guys used Pydantic. And I assumed there was going to be a good one that I could go and use. Started looking at them and was super disappointed by what I found."
Not counted
Samuel Colvin founded Pydantic, whose Pydantic AI framework competes with CrewAI. Kept in the record; it takes part in no figure.
No opinion stated
How can disable tracing prompt?GitHub issue
A tracing prompt added about 20 seconds to each run and kept appearing after setting tracing=False and CREWAI_TRACING_ENABLED=false.
Counts toward, pending a recheck
Negative
Comment on: Ask HN, what are you working on (September 2025)?Hacker News comment
Started with CrewAI, LangGraph, and others, found the learning curve more than needed for the features used, and built tinyAgent instead.
No opinion stated
Agent does not actually invoke tools, only simulates tool usage with fabricated outputGitHub issue
The agent wrote a plausible Thought, Action, Observation trace while never running the search tool; the model invented the observation.
Counts toward, pending a recheck
No opinion stated
Even though I turned off the telemetry feature, telemetry is still being sentGitHub issue
Set CREWAI_DISABLE_TELEMETRY to true as the docs describe and still saw telemetry sent.
Counts toward, pending a recheck
No opinion stated
Tool Called Multiple Times in CrewAI v0.121.0GitHub issue
After upgrading to 0.121.0, a tool ran several times within one task although the first call succeeded; 0.120.0 called it once.
Negative
Comment on: Show HN, Cleverb.ee, an open-source agent that writes a cited research reportHacker News comment
Rebuilt the research tool with CrewAI but found it limiting, and was rebuilding it again on AutoGen.
Positive
CrewAI, Technology Radar (Assess)Company technology publication
Thoughtworks placed CrewAI in Assess after using it several times on production problems, from validating promotion codes to investigating transaction failures and customer support queries.
Positive
Comment on: Ask HN, open-source AI agents building platform?Hacker News comment
Runs the crewAI library with ChromaDB and local Ollama models, and had to customize CrewAI's RAG tool to work with them; likes the class setup for building fully custom tools.
Pages from CrewAI, used to check facts for every group.
Describes Flows as event-driven workflows that manage state and control execution, and Crews as teams of autonomous agents that collaborate on tasks a Flow delegates.
Documents start, listen, and router steps and the @persist decorator, which saves flow state to SQLite by default so a flow can continue across restarts.
Documents checkpoints that let a crew, flow, or agent resume after a failure or fork into a new branch. On restore, completed tasks are skipped.
Models are set as provider/model-id in the MODEL variable, agent YAML, or code. Native SDK integrations cover OpenAI, Anthropic, Gemini, Azure, Bedrock, and Snowflake Cortex, and every other provider runs through LiteLLM.
memory=True gives a crew shared memory, which uses OpenAI text-embedding-3-large embeddings unless a custom embedder is configured.
CrewAI collects anonymous telemetry by default, with more detail under share_crew. CREWAI_DISABLE_TELEMETRY or OTEL_SDK_DISABLED turns it off, and AMP tracing needs CREWAI_TRACING_ENABLED=false separately.
The @human_feedback decorator, from CrewAI 1.8.0, pauses a flow for a person's review and routes on the answer.
Basic is free with 2 automations and 50 workflow executions a month as a hard maximum. Enterprise is custom, with included executions sized to the workflow, flexible overage, SSO with Microsoft Entra or Okta, and deployment on CrewAI cloud, a dedicated VPC, or the customer's infrastructure.
CrewAI AMP deploys crews and flows to managed infrastructure, exposes them through a REST API, and traces each execution.
AMP routes Flow review requests by email and dashboard and lists immutable audit logs of review decisions, with reviewer identity and timestamps.
MIT-licensed Python framework requiring Python 3.10 to 3.13, with 60,218 stars on October 28, 2026. The README describes the anonymous telemetry and OTEL_SDK_DISABLED.
The crewai package requires chromadb~=1.5.9 on October 28, 2026. GitHub advisory GHSA-f4j7-r4q5-qw2c, published May 18, 2026, rates ChromaDB 1.0.0 to 1.5.9 critical for pre-authentication code injection and lists no patched version.
DEFAULT_LLM_MODEL is gpt-4.1-mini, used when MODEL, MODEL_NAME and OPENAI_MODEL_NAME are unset. The Agents docs still name gpt-4 as the default.
crewai 1.15.32 was released on October 27, 2026, the second release that day and the fifth between October 14 and October 27.
Source index
Every source GILD read, with the original address and the context used to count it. Open a finding to filter the full cards above. On smaller screens the index lists each source once by name, cohort and date; its full record above carries the address, context and findings.
| Source | Original URL | Author and context | Who | Published | Review date | Applies | Findings |
|---|---|---|---|---|---|---|---|
| Two agents set up at the same time can fail with RuntimeError: dictionary changed size during iterationID gh-8697 | Developer Communities | github.com/ | Yifan Xiong (yifanxiong272), GitHub user filing an issueself-described: Flows instantiated concurrently in one process | Individual builder | Published Oct 28, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| A failed task can be skipped after checkpoint restore because its output remains setID gh-8718 | Developer Communities | github.com/ | ssshr-66, GitHub user filing an issueself-described: Checkpoints triggered on task failure | Individual builder | Published Oct 28, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| [chore]: rm crewai integrationID sf-7 | Developer Communities | github.com/ | Sean McGuire, Browserbase (pull request author)self-described: Stagehand, Browserbase's browser automation framework, and its integrations | Product or platform team | Published Oct 27, 2026 | Assessment reviewed Oct 30, 2026 | Current | Listed, no finding position |
| Bedrock prompt_tokens excludes cache reads and writes, unlike other native providersID gh-8626 | Developer Communities | github.com/ | r14dd, GitHub user filing an issueself-described: Comparing token counts across CrewAI's native providers | Individual builder | Published Oct 26, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Post on once, an idempotency library for agent tool callsID x-frogscooper | Developer Communities | x.com/ | frogscooper, AI engineer posting on Xself-described: Built a library that runs each agent tool call at most once per key | Individual builder | Published Oct 25, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| UploadCache sync wrappers deadlock 30s inside a running event loopID gh-8516 | Developer Communities | github.com/ | Wilson0511, GitHub user filing an issueself-described: File uploads from async apps and Jupyter on crewai-files 1.15.30 | Individual builder | Published Oct 24, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| I built an open-source app where LLM agents debate a stock before a manager agent decides the tradeID rd-1xea1cs | Developer Communities | www.reddit.com/ | kingabzpro, Reddit poster in r/AI_Agentsself-described: BetterTradingAgents, an open-source stock research app on FastAPI and CrewAI | Individual builder | Published Oct 24, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Post on building a research and writing crewID x-emagzimus | Developer Communities | x.com/ | Emmanuel Chukwu (Emagzimus), builder posting on Xself-described: A research agent and a content agent that draft blog posts | Individual builder | Published Oct 8, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Agentic AI Frameworks Compared 2026: LangGraph WinsID sf-11 | Engineering Blogs & Reviews | dev.to/ | Shaam (shaam_ai), dev.to authorself-described: Scoring five agent frameworks on control, license and production use | Individual builder | Published Oct 2, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| deps!: drop the crewai extra to remove chromadb from the lockfileID sf-4 | Developer Communities | github.com/ | Tejas Chopra, Headroom maintainer (pull request author)self-described: Headroom, an open-source project with a CrewAI tool adapter | Small product team | Published Sep 30, 2026 | Assessment reviewed Oct 30, 2026 | Current | Listed, no finding position |
| Ran one build on eight agent platforms. Two finished.ID rd-1wpwbb9 | Developer Communities | www.reddit.com/ | yiddo_bhushan, operations at Shogo, one of the eight platforms testedself-described: One document question-and-answer job run on eight agent platforms by a non-developer | Individual builder | Published Sep 25, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| [BUG] GPT-6 Astra agents with tools fail on Bedrock's OpenAI-compatible API because the retry sends reasoning_effort 'none'ID gh-7740 | Developer Communities | github.com/ | Lucas Kim, AI/ML engineer and consultant whose GitHub profile lists AWSself-described: CrewAI agents with tools on GPT-6 Astra, GPT-6 Sol, and GPT-6 Luna through Bedrock's OpenAI-compatible endpoints | Individual builder | Published Sep 24, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Comment on: LangGraph or CrewAI for agent handoffsID rd-pbdjg39 | Developer Communities | www.reddit.com/ | Apart-Currency7380, Reddit commenter in r/LangChainself-described: Choosing a framework where handoffs between agents were the pain point | Individual builder | Published Sep 22, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| I built the same agent in Strands, LangGraph, and CrewAI and recorded every LLM callID dev-sunnydachs | Engineering Blogs & Reviews | dev.to/ | sunnydachs, developer writing on dev.toself-described: 27 recorded runs of one agent across three frameworks | Individual builder | Published Sep 21, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| LangGraph vs CrewAI vs OpenAI Agents SDK: Picking Your Agent Framework in 2026ID sf-9 | Engineering Blogs & Reviews | dev.to/ | jamilxt, dev.to authorself-described: A researched comparison of three agent frameworks, built from documentation and published reports | Individual builder | Published Sep 15, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| ToolUsage retries a failed tool call twice per outer attempt, not onceID gh-7449 | Developer Communities | github.com/ | giskard09, GitHub user filing an issueself-described: Counting retries on failing tool calls | Individual builder | Published Sep 14, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| LanceDBStorage.delete() ignores scope/older_than filters and causes accidental mass deletionID gh-7419 | Developer Communities | github.com/ | Rohit Kanithi (Rohitkanithi), GitHub user filing an issueself-described: Scoped, multi-tenant memory on the default LanceDB store | Individual builder | Published Sep 12, 2026 | Assessment reviewed Oct 30, 2026 | Needs a recheck | |
| Comment on: An Accidental BlackboardID hn-49625124 | Developer Communities | news.ycombinator.com/ | windexh8er, Hacker News commenterself-described: About two years of watching agents collaborate in CrewAI | Individual builder | Published Sep 9, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Memoized CrewBase methods retain per-instance Agent/Task/Crew in global cacheID gh-7338 | Developer Communities | github.com/ | suqinghen, GitHub user filing an issueself-described: A new crew instance created per request in a service | Individual builder | Published Sep 8, 2026 | Assessment reviewed Oct 30, 2026 | Needs a recheck | |
| Post on graph engineering in production multi-agent buildsID x-baglieri | Developer Communities | x.com/ | Carlos Matias Baglieri, founder of FACTAself-described: Production multi-agent builds on CrewAI, LangGraph, and Agno | Technical leadership | Published Sep 6, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| crew.usage_metrics multiplies token usage by the number of agents sharing an LLM instanceID gh-7259 | Developer Communities | github.com/ | parthiban-sivakumar, GitHub user filing an issueself-described: Several agents sharing one LLM object | Individual builder | Published Sep 4, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| I delegated a review in CrewAI. The reviewer ran an export tool the orchestrator never had.ID dev-rafael | Engineering Blogs & Reviews | dev.to/ | Rafael Asor, author of the attenu-guard tool-gating libraryself-described: Delegation test on crewai 1.15.18 with a scripted LLM and recorded tool calls | Individual builder | Published Sep 3, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Agent.kickoff_async guardrail retry calls sync invoke inside event loopID gh-7252 | Developer Communities | github.com/ | Jucelio Quentino (jquentino-ambev), engineer at AB InBev filing an issueself-described: Agents run with kickoff_async and output guardrails | Product or platform team | Published Sep 3, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| refactor: delete vendored CrewAi and mem0 librariesID sf-6 | Developer Communities | github.com/ | Vitalii Povoroznyk, EpicStaff contributor (pull request author)self-described: EpicStaff, a self-hosted platform for building agent flows over a Django backend, removing CrewAI from its own code | Product or platform team | Published Aug 20, 2026 | Assessment reviewed Oct 30, 2026 | Current | Not counted. EpicStaff builds its own agent-flow platform, an alternative to CrewAI's. |
| Truncated responses are detected in the Bedrock provider onlyID gh-7013 | Developer Communities | github.com/ | Joaquin DG (JoaquinDG), GitHub user filing an issueself-described: Reproduced against crewai 1.15.16 across native providers | Individual builder | Published Aug 17, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| AI Agent Frameworks Compared 2026: LangGraph vs CrewAI vs OpenAI SDKID sf-10 | Engineering Blogs & Reviews | dev.to/ | Jangwook Kim, dev.to authorself-described: Comparing six agent frameworks for production | Individual builder | Published Aug 9, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| CrewAI's quadratic context problem: why a 5-agent crew costs 6x more than you expectID dev-wartzarbee | Engineering Blogs & Reviews | dev.to/ | wartzarbee, developer writing a cost-audit series on dev.toself-described: Token cost of a sequential five-agent crew with default context passing | Individual builder | Published Jul 26, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Why I'm Betting on CrewAI for Multi-Agent Orchestration (And Where It Falls Short)ID dev-ashg2099 | Engineering Blogs & Reviews | dev.to/ | ashg2099, developer writing on dev.toself-described: Technical review of CrewAI for research and content pipelines | Individual builder | Published Jul 8, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Remove crewai to fix CRITICAL chromadb vulnerabilityID sf-5 | Developer Communities | github.com/ | Chisanan232, ai-agent-assembly maintainer (pull request author)self-described: Framework smoke tests for ai-agent-assembly's end-to-end suite | Small product team | Published Jul 7, 2026 | Assessment reviewed Oct 30, 2026 | Current | Listed, no finding position |
| Tool exceptions are caught and replaced with generic error messagesID gh-6262 | Developer Communities | github.com/ | Thomas Connally (tcconnally), GitHub user filing an issueself-described: Debugging failed tool calls | Individual builder | Published Jun 20, 2026 | Assessment reviewed Oct 30, 2026 | Needs a recheck | |
| Comment on: Show HN, RiskKernel, kill -9 an AI agent and resume it without paying twiceID hn-48476114 | Developer Communities | news.ycombinator.com/ | stockanalyst987, Hacker News commenterself-described: An agentic system built on CrewAI | Individual builder | Published Jun 10, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Comment on: building agent workflows across coding agentsID sf-1 | Developer Communities | news.ycombinator.com/ | matbanik, Hacker News commenterself-described: Runs agent workflows for business on the API and on subscriptions for personal work | Individual builder | Published Jun 1, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| input_files (PDFFile) are passed as base64 via read_file tool, causing context overflowID gh-5930 | Developer Communities | github.com/ | axelgbl, GitHub user filing an issueself-described: Crews that take PDF inputs | Individual builder | Published May 26, 2026 | Assessment reviewed Oct 30, 2026 | Needs a recheck | |
| cache_breakpoint injected into messages for non-Anthropic providersID gh-5886 | Developer Communities | github.com/ | Hamza-BinRiaz, GitHub user filing an issueself-described: Agents on Groq and OpenAI-compatible endpoints | Individual builder | Published May 21, 2026 | Assessment reviewed Oct 30, 2026 | Needs a recheck | |
| Tool re-execution on task retry has no idempotency guardID gh-5802 | Developer Communities | github.com/ | azender1, GitHub user filing an issueself-described: Tools that call payment, email, or trading APIs | Individual builder | Published May 14, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Comment on: How many products does Microsoft have named 'Copilot'?ID hn-47652430 | Developer Communities | news.ycombinator.com/ | dsteel, Hacker News commenter whose team runs 14 AI agentsself-described: A team running 14 AI agents | Small product team | Published Apr 5, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Comment on: Show HN, I tested 11 AI frameworks for basic securityID hn-47616145 | Developer Communities | news.ycombinator.com/ | Acacian, Hacker News commenter and author of the aegis guardrail libraryself-described: Reviewed 11 frameworks for injection detection, PII masking, and audit trails | Individual builder | Published Apr 2, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Comment on: Tell HN, Litellm 1.82.7 and 1.82.8 on PyPI are compromisedID hn-47502715 | Developer Communities | news.ycombinator.com/ | eoskx, Hacker News commenterself-described: Dependency exposure of agent frameworks during the March 2026 LiteLLM compromise | Individual builder | Published Mar 24, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Comment on: Tell HN, Litellm 1.82.7 and 1.82.8 on PyPI are compromisedID hn-47510350 | Developer Communities | news.ycombinator.com/ | mathis-l, Hacker News commenterself-described: CrewAI's response to the March 2026 LiteLLM 1.82.7 and 1.82.8 compromise | Individual builder | Published Mar 24, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| One Chat App, 4 AI Frameworks ComparedID blog-vstorm | Engineering Blogs & Reviews | oss.vstorm.co/ | Kacper Wlodarczyk, AI engineer at Vstormself-described: The same FastAPI and Next.js chat app built on Pydantic AI, LangChain, LangGraph, and CrewAI | Product or platform team | Published Mar 13, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Why Checkpoints Aren't Durable Execution: LangGraph, CrewAI, Google ADK, and OthersID blog-diagrid | Engineering Blogs & Reviews | www.diagrid.io/ | Yaron Schneider, CTO and co-founder of Diagrid, which sells a durable workflow runtimeself-described: Recovery of multi-step agent workflows after a process crash | Technical leadership | Published Feb 25, 2026 | Assessment reviewed Oct 30, 2026 | Current | Not counted. Diagrid sells a durable workflow runtime that the post recommends over CrewAI persistence. |
| Comment on: Does anyone use CrewAI or LangChain anymore?ID hn-47132417 | Developer Communities | news.ycombinator.com/ | obiefernandez, Hacker News commenterself-described: Reply to a February 2026 thread asking who still uses CrewAI or LangChain | Individual builder | Published Feb 24, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Comment on: Show HN, Intellistant, a C++ alternative to LangChain for AI agentsID hn-46511965 | Developer Communities | news.ycombinator.com/ | pooriayousefi, author of Intellistant, a C++ multi-agent frameworkself-described: Long-running agent workflows on local LLMs | Individual builder | Published Jan 6, 2026 | Assessment reviewed Oct 30, 2026 | Current | |
| Token tracking race condition in threading-based async_executionID gh-4168 | Developer Communities | github.com/ | Devasy, GitHub user filing an issueself-described: Tasks with async_execution=True on one agent | Individual builder | Published Jan 3, 2026 | Assessment reviewed Oct 30, 2026 | Needs a recheck | |
| Comment on: Critical vulnerability in LangChain, CVE-2025-68664ID hn-46393633 | Developer Communities | news.ycombinator.com/ | alzoid, Hacker News commenterself-described: Evaluated LangChain, AG2, Google's offerings, CrewAI, and Microsoft's frameworks | Individual builder | Published Dec 26, 2025 | Assessment reviewed Oct 30, 2026 | Current | |
| Comment on: I added native time awareness to CrewAI to fix LLM date hallucinationsID hn-46264408 | Developer Communities | news.ycombinator.com/ | sherwin27, Hacker News commenter who contributed tools to CrewAIself-described: AI agents in production that hallucinated the current date | Individual builder | Published Dec 14, 2025 | Assessment reviewed Oct 30, 2026 | Current | |
| Pydantic AI with Samuel ColvinID sf-8 | Podcasts, Talks & Videos | softwareengineeringdaily.com/ | Samuel Colvin, founder of Pydanticself-described: Choosing a framework for AI features in Pydantic's Logfire before building Pydantic AI | Technical leadership | Published Dec 4, 2025 | Assessment reviewed Oct 30, 2026 | Current | Not counted. Samuel Colvin founded Pydantic, whose Pydantic AI framework competes with CrewAI. |
| How can disable tracing prompt?ID gh-3789 | Developer Communities | github.com/ | IbrahimCanKALYA, GitHub user filing an issueself-described: Crews run repeatedly from a script | Individual builder | Published Oct 24, 2025 | Assessment reviewed Oct 30, 2026 | Needs a recheck | |
| Comment on: Ask HN, what are you working on (September 2025)?ID hn-45376520 | Developer Communities | news.ycombinator.com/ | mahdiyar, Hacker News commenter building tinyAgentself-described: A year of trying agent frameworks before writing a small framework of their own | Individual builder | Published Sep 25, 2025 | Assessment reviewed Oct 30, 2026 | Current | |
| Agent does not actually invoke tools, only simulates tool usage with fabricated outputID gh-3154 | Developer Communities | github.com/ | redvelvets, GitHub user filing an issueself-described: Custom LLM with a web search tool and Phoenix memory | Individual builder | Published Jul 14, 2025 | Assessment reviewed Oct 30, 2026 | Needs a recheck | |
| Even though I turned off the telemetry feature, telemetry is still being sentID gh-2945 | Developer Communities | github.com/ | fzozyurt, GitHub user filing an issueself-described: Crew created with crewai create, telemetry disabled in main.py | Individual builder | Published Jun 3, 2025 | Assessment reviewed Oct 30, 2026 | Needs a recheck | |
| Tool Called Multiple Times in CrewAI v0.121.0ID gh-2881 | Developer Communities | github.com/ | DipakBundheliya, GitHub user filing an issueself-described: Agent with one tool that worked on 0.120.0 | Individual builder | Published May 22, 2025 | Assessment reviewed Oct 30, 2026 | Needs a recheck | |
| Comment on: Show HN, Cleverb.ee, an open-source agent that writes a cited research reportID hn-43877178 | Developer Communities | news.ycombinator.com/ | nickwatson, Hacker News commenter building a research toolself-described: Multi-agent research tool with critique and fact-check roles | Individual builder | Published May 3, 2025 | Assessment reviewed Oct 30, 2026 | Current | |
| CrewAI, Technology Radar (Assess)ID blog-thoughtworks | Engineering Blogs & Reviews | www.thoughtworks.com/ | Thoughtworks, Technology Radar volume 32self-described: Thoughtworks teams using CrewAI on client production problems | Product or platform team | Published Apr 2, 2025 | Assessment reviewed Oct 30, 2026 | Needs a recheck | |
| Comment on: Ask HN, open-source AI agents building platform?ID hn-42635726 | Developer Communities | news.ycombinator.com/ | dabe19, Hacker News commenterself-described: Local RAG setup on a personal PC with an RTX 3060, ChromaDB, and Ollama models | Individual builder | Published Jan 8, 2025 | Assessment reviewed Oct 30, 2026 | Current | |
| CrewAI introductionID cd-01 | Product Documentation | docs.crewai.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| FlowsID cd-02 | Product Documentation | docs.crewai.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| CheckpointingID cd-03 | Product Documentation | docs.crewai.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| LLMsID cd-04 | Product Documentation | docs.crewai.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| MemoryID cd-05 | Product Documentation | docs.crewai.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| TelemetryID cd-06 | Product Documentation | docs.crewai.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| Human feedback in FlowsID cd-07 | Product Documentation | docs.crewai.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| CrewAI pricingID cd-08 | Product Documentation | www.crewai.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| CrewAI AMP introductionID cd-09 | Product Documentation | docs-platform.crewai.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| Flow HITL ManagementID cd-10 | Product Documentation | docs-platform.crewai.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| CrewAI on GitHubID cd-11 | Product Documentation | github.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| CrewAI source: pyproject.toml dependenciesID cd-12 | Product Documentation | github.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| CrewAI source: constants.py default modelID cd-13 | Product Documentation | github.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
| CrewAI releases on GitHubID cd-14 | Product Documentation | github.com/ | CrewAIOfficial CrewAI documentation, pricing, and repository pages | Not counted (documentation) | No date stated | Documentation checked Oct 28, 2026 | Current | Verification only |
