Competitive landscape¶
This is a buyer's comparison, not a claim that every product solves the same problem. Each row says what the linked project actually offers and where its boundary differs from Fgentic. Every factual observation carries its own checked date; a later product or repository change can invalidate it. Repository popularity is a point-in-time adoption signal, not a quality score.
Head-to-head¶
| Project | What is available now | Boundary relative to Fgentic |
|---|---|---|
| AgentTeams | The Apache-2.0 project offers a one-command local installer that bundles Matrix, Element, an AI gateway, shared storage and a manager/worker runtime, plus a Helm-based Kubernetes install; setup still requires an LLM provider and API key. Its current v1.2.0 stable release adds an optional dashboard for visual resource and Matrix management. (README at audited revision, v1.2.0 release, license at audited revision; checked 2026-08-01) |
It is the closest direct alternative and wins on integrated-runtime onboarding. Its README advertises optional Matrix federation, but the audited first-party architecture and install evidence cover one bundled Tuwunel homeserver; that is not proof that it cannot federate, nor cross-homeserver acceptance evidence. Its runtime contract is native manager/worker orchestration; Fgentic instead proves a closed federation profile and makes A2A the pinned boundary to independently replaceable agents. (federation statement, default topology, Fgentic federation contract; checked 2026-08-01) |
| Buzz | The Apache-2.0 project ships a self-hostable Nostr workspace with channels, threads, DMs, canvases, media, search, an audit log, an agent-first CLI, YAML workflows and Git events under a signed-event model. The current single-relay setup maps one relay URL to one community. Each agent has its own Nostr keypair; Buzz reveals the secret at creation for the operator to save, and reusing that secret preserves the same pubkey identity. (README at audited revision, shipped feature table, agent key contract, license at audited revision; checked 2026-08-01) | Buzz wins on portable per-agent key identity and on integrating chat, agents, code review, workflows and Git activity into one signed event model. The keypair can be reused; memberships, history, DMs and community state remain community-local. Its README lists cross-relay web-of-trust as pending, whereas Fgentic's current boundary is closed Matrix federation plus explicitly pinned remote A2A routes; Fgentic does not provide one agent-held portable workspace key or an integrated software forge. (Buzz community boundary, Buzz status table, Fgentic federation contract; checked 2026-08-01) |
| Slack Agentforce | Agentforce agents work in Slack channels, DMs and threads; users can @mention them, administrators add them to an Agents tab, and templates provide guided setup. Agents can use permitted Slack and Salesforce context and take Slack actions such as creating channels, updating canvases and sending DMs. Use is available on Slack plans only with a Salesforce Agentforce license. (product page, user guide; checked 2026-08-01) |
For an existing Slack-and-Salesforce organization, Slack Agentforce wins on managed onboarding, agent discovery, familiar workspace UX and immediate access to enterprise conversation context. Its identity, authorization, data and license boundaries are Slack and Salesforce tenant services; Fgentic trades that convenience for an operator-owned open-source Matrix, GitOps, model and federation control plane. (Slack Agentforce product page, Slack user guide, Fgentic production path, Fgentic exit strategy; checked 2026-08-01) |
| Dify | Workflow Studio provides a collaborative visual canvas for model calls, retrieval, tools, code, branches, triggers and human review. Chatflow conversation variables persist across turns within one chat session. A paused Human Input run continues after its form token is submitted; separately, clients can reconnect the event stream by workflow run ID and replay executed-node state. Dify offers cloud and self-hosted deployment, under a modified Apache-2.0 license with additional multi-tenant and frontend-branding conditions. (Workflow Studio, conversation variables, human-input flow, streaming contract, license at audited revision; checked 2026-08-01) | Dify wins on visual workflow authoring, session-scoped conversation state and operator-facing resumable execution. Those documented pause/reconnect semantics do not establish the same crash-recovery guarantee as Fgentic's Matrix-to-A2A ledger. Dify is an AI-application builder rather than a federated shared-room identity fabric; Fgentic is narrower and has no comparable visual builder. (Dify Workflow Studio, Dify human-input flow, Fgentic durable ledger; checked 2026-08-01) |
| Nextcloud Assistant | Assistant is the UI for Nextcloud's modular AI task API and chat, including summarization, transcription, image generation and context-aware answers; an administrator selects backend apps and may run supported workloads on-premises. (administrator manual; checked 2026-07-31) | It is a strong fit when the collaboration and knowledge boundary is already Nextcloud. Fgentic instead treats federated Matrix rooms as the human collaboration plane and delegates agent tasks over A2A. (Nextcloud administrator manual, Fgentic architecture; checked 2026-07-31) |
| Nextcloud Context Agent | Context Agent adds action tools for Nextcloud apps and exposes all of those tools through an MCP server for third-party agents. It therefore already ships an open agent protocol; the distinction is not “Fgentic has protocols and Nextcloud does not.” (administrator manual, source repository; checked 2026-07-31) | Context Agent's tool and authorization scope is a Nextcloud installation. Fgentic composes MCP-capable agents behind an A2A delegation boundary and adds room, sender and cross-organization admission policy. (Nextcloud administrator manual, Fgentic architecture, Fgentic security boundary; checked 2026-07-31) |
| Mattermost Agents | The plugin supplies multiple configurable assistants, channel summaries, permissions, and local or hosted LLM choices inside Mattermost. Most of the repository is Apache-2.0, with server/enterprise under Mattermost's source-available license. (README and license notice; checked 2026-07-31) |
It has the lower-friction path for an existing Mattermost workspace and a more integrated admin workflow. Fgentic is a separate, federated Matrix collaboration fabric whose agents are replaceable A2A services rather than one chat product's plugin. (Mattermost README, Fgentic coexistence guide; checked 2026-07-31) |
| Rocket.Chat AI | Rocket.Chat AI provides self-hosted retrieval, semantic search, grounded answers, summaries and an AI bot that users can mention in channels, while enforcing existing Rocket.Chat access controls. (product documentation; checked 2026-07-31) | It is already useful for governed knowledge work in a Rocket.Chat deployment. Fgentic's narrower present focus is portable A2A delegation and cross-homeserver collaboration, not a more mature in-product search assistant. (Rocket.Chat documentation, Fgentic architecture; checked 2026-07-31) |
| MindRoom | The Apache-2.0 Matrix-native runtime gives agents Matrix accounts, multi-agent threads, persistent memory, more than 100 tool integrations, RAG, model routing and a web dashboard; it supports self-hosted or hosted Matrix and explicitly describes federation. (README, repository metadata; checked 2026-07-31) | MindRoom currently offers a broader end-user agent experience and tool catalog. Fgentic separates the Matrix appservice from agent implementations through A2A and concentrates on Kubernetes/GitOps governance and pinned cross-organization delegation. (MindRoom README, Fgentic bridge contract, Fgentic production path; checked 2026-07-31) |
| AGNTCY | AGNTCY is an ecosystem of composable infrastructure rather than a team-chat product: its repositories cover agent schemas, distributed discovery and secure low-latency messaging. (organization repositories, documentation; checked 2026-07-31) | It is complementary upstream infrastructure. Fgentic uses Matrix for visible human collaboration, and has adopted AGNTCY-compatible identity and discovery direction without replacing the room or A2A task layers. (AGNTCY documentation, Fgentic ADR 0022; checked 2026-07-31) |
| Microsoft Copilot Studio (multitenant) | Copilot Studio is a low-code managed agent platform whose capacity, request quotas and billing are allocated through Microsoft tenants and Dataverse environments. (Microsoft quotas and limits, capacity administration; checked 2026-07-31) | It provides a substantially more complete builder and enterprise administration surface, including tenant-scale commercial operations. Its control and billing plane remains Microsoft-managed; Fgentic trades that convenience for an operator-owned Kubernetes, identity, model and federation boundary. (Microsoft capacity administration, Fgentic production path, Fgentic exit strategy; checked 2026-07-31) |
| Gemini Enterprise Agent Registry | Gemini Enterprise presents a managed platform for discovering and using Google, partner and custom agents, with centralized governance and an agent marketplace/registry experience. (product page, publishing guide; checked 2026-07-31) | It offers a far richer managed catalog and distribution path. Its registry and control plane are Google Cloud services; Fgentic's agent routes and trust pins live in adopter-controlled Git and clusters, with no hosted catalog required for the core path. (Gemini Enterprise product page, Fgentic production path, Fgentic federation contract; checked 2026-07-31) |
kagent a2a-slack-template |
The small template connects a Slack bot to kagent over A2A and links a walkthrough; it demonstrates that a direct chat-to-A2A adapter can be much smaller than a collaboration platform. (template README; checked 2026-07-31) | It is the simpler choice for one Slack workspace and one integration. Fgentic takes on the additional cost of a sovereign Matrix plane, sender/room policy, durable task handling and federated trust controls. (template README, Fgentic bridge contract, Fgentic federation contract; checked 2026-07-31) |
Fgentic's defensible differentiators¶
These are current, evidence-backed boundaries—not universal advantages over every project above:
| Decision axis | Current Fgentic evidence |
|---|---|
| Cross-organization trust | Closed room-v12 Matrix federation, server ACLs and sender policy govern the shared-room plane; verified Signed AgentCards plus machine credentials and route pins govern remote A2A delegation. The permanent provider-free rig exercises both planes. (federation contract; checked 2026-08-01) |
| Operator delivery and cost guard | Flux and SOPS keep delivery and secrets under adopter control. The bridge rate-limits per sender and room; agentgateway reports aggregate model-token activity and cost-catalog lookup coverage; cross-org maxTokens is labeled as an admission reservation rather than measured spend. (production path, cost dashboards, federation cost boundary; checked 2026-08-01) |
| Stewardship and license honesty | Fgentic is an Apache-2.0 composition of independently stewarded open protocols and runtimes. No single foundation governs, operates, certifies or endorses the complete runtime, and no upstream foundation status transfers to Fgentic. (open-stack governance map, licensing map; checked 2026-08-01) |
Where competitors win today¶
These are current disadvantages, not items hidden behind a future roadmap:
| They win on | Competitor | Evidence and Fgentic boundary |
|---|---|---|
| Onboarding speed | AgentTeams | Its one-command local installer bundles the room, gateway, storage and runtime, although it still asks for an LLM provider and API key. Fgentic's credential-free local demo reduces evaluation cost, but its production path requires Kubernetes, Flux, SOPS, DNS, identity and several acceptance gates. (AgentTeams install, Fgentic evaluator, Fgentic production installation; checked 2026-08-01) |
| Portable agent identity | Buzz | Each Buzz agent has its own Nostr keypair; the operator must save the revealed secret, and reusing it preserves the pubkey workspace identity. Memberships, history, DMs and community state remain local. Fgentic's ghost identity is homeserver-governed and its separate remote-agent identity is an operator-pinned Signed AgentCard, not one portable agent-held workspace key. (Buzz agent key contract, Buzz community boundary, Fgentic federation contract; checked 2026-08-01) |
| Managed workspace adoption | Slack Agentforce | For an existing Slack-and-Salesforce organization, admin-added agents, an Agents tab, guided templates, channel mentions and existing permissions substantially reduce deployment and user-training work. Fgentic asks the adopter to operate the collaboration and agent control planes. (Slack Agentforce, Slack user guide, Fgentic production path; checked 2026-08-01) |
| Visual building and session memory | Dify | Dify combines a visual node canvas with conversation variables that persist across turns within a chat session, plus debugging, error branches, traces and version restore. Fgentic exposes typed source and GitOps contracts and has no comparable visual application builder. (Dify Workflow Studio, Dify conversation variables, Fgentic extending guide; checked 2026-08-01) |
| Resumable workflow operations | Dify | Dify documents form-token continuation after human input, plus workflow-run-ID event-stream reconnection and executed-node state replay. That is a more complete operator-facing workflow lifecycle than Fgentic's bridge-focused surface; it is not evidence that both products make identical process-crash guarantees. (Dify human-input flow, Dify streaming contract, Fgentic durable ledger; checked 2026-08-01) |
| Admin surface and ready-made usefulness | MindRoom and others | MindRoom has a web dashboard, persistent memory and more than 100 tool integrations; Nextcloud Context Agent can act across the Nextcloud suite; Rocket.Chat AI ships search and grounded workspace answers. Fgentic's optional Ketesa console is intentionally bounded, and its checked-in sample-agent gate proves a deterministic source contract rather than useful live-model answers. (MindRoom README, Nextcloud Context Agent, Rocket.Chat AI, Fgentic admin console, model evidence boundary; checked 2026-07-31) |
| Community discoverability | AgentTeams | The public GitHub snapshot showed 5,247 stars for AgentTeams and 2 for Fgentic—roughly a 2,600-fold difference. Stars do not prove production adoption, but the discoverability and contributor gap is real. (AgentTeams repository metadata, Fgentic repository metadata; checked 2026-07-31) |
Durable-execution boundary. Dify's cited primary sources support human-wait continuation and event-stream reconnection, not a general process-crash or exactly-once guarantee. This page therefore awards Dify the narrower resumable-operations win instead of relabeling it as crash-durable execution. (Dify human-input flow, Dify streaming contract, Fgentic durable ledger; checked 2026-08-01)
Why not AgentTeams?¶
Choose Fgentic instead of AgentTeams when the non-negotiable requirement is an independently replaceable A2A agent boundary with pinned, authenticated cross-organization delegation; choose AgentTeams when its integrated manager/worker runtime and materially easier installation matter more than that boundary. (AgentTeams README at audited revision, Fgentic bridge contract, Fgentic federation contract; checked 2026-08-01)
That sentence is deliberately conditional: AgentTeams is open source, self-hostable, Matrix-based and Kubernetes-capable, and its README advertises optional federation. “Sovereignty” or “open standards” alone is not an honest separator; the narrower distinction is Fgentic's executable cross-homeserver trust and A2A boundary versus AgentTeams' integrated runtime and currently documented one-homeserver topology. (AgentTeams federation statement, AgentTeams topology, AgentTeams license; checked 2026-08-01)
Refresh cadence¶
Refresh this page during preparation for the first Friday TWiM entry in January, April, July and October:
- Reopen every linked primary source, starting with AgentTeams, Buzz, Slack Agentforce and Dify.
- Recheck each “available now,” competitor-win, topology and license claim; distinguish shipped behavior from a vision or roadmap.
- Update the checked date only on claims actually revalidated, and remove claims whose source no longer supports them.
- Review new direct alternatives and remove rows that no longer help an adopter decide.
- Treat a material comparison change as a TWiM candidate only when it produces a watchable artifact; otherwise follow the TWiM runbook and do not manufacture an announcement.
Issue #1617 owns the initial external-publication evidence. The repository owns this quarterly source check even when no TWiM entry is eligible.