Team Agent Platforms

Related: For individual AI assistants (email, calendar, personal productivity), see Personal Agents Platforms Compared. For coding-first agent platforms, see Cloud Coding Agent Platforms and Foundation Lab Coding Agents.

Review dates: Agent Studio's coverage and the adjacent Grok Bot discussion were checked September 16, 2026. Other vendor entries retain their June 11 review baseline; their prices, availability, and market figures have not been reverified in this targeted update.

Executive Summary

Team agent platforms deploy shared AI agents that entire teams can access via Slack, Discord, or APIs. Unlike personal AI assistants (managing your email and calendar) or local IDE tools (your personal Cursor or Copilot), these platforms enable collaborative, asynchronous workflows across functions — from product planning and data analysis to customer support and operations.

This report focuses on platforms whose primary job is general team work. Platforms that are primarily software factories or coding agent platforms (Devin, Tembo, Codex, Kilo, eksec) were moved out of this category in the June 2026 revision — they're covered in the coding-agent reports linked above. Spacebot, initially swept out with them, returns this cycle: it is a general-purpose community/team agent framework, not a coding platform.

Key Findings:

• Capital is flooding in — Viktor raised a $75M Series A led by Accel with a $15M run-rate three months post-launch; Dust raised $40M from Abstract and Sequoia; Glean hit a $7.2B valuation^[1][2][3]
• "Multiplayer AI" is the new framing — Dust's 3,000+ orgs run 300K+ shared agents on company data; Glean reports 250M+ agent actions over its indexed knowledge^[2][3]
• Claude leads in breadth — cowork in Slack spans product planning, research, coding, and writing — the widest capability set^[4][5]
• Agent definitions and runtime review are separate decisions — Tembo Agent Studio stores definitions in Git, defaults authoring to PRs, and provides a shared Tasks Inbox for reviewing proposed work. Direct-commit mode is optional; self-hosting does not remove connected services from the data path.^[6][7][8]
• The platform owner is now a competitor — Slack is rebuilding Slackbot as a first-party AI teammate, a channel risk for every vendor here^[9]
• Adoption is mainstream — 72% of enterprises are using or testing agents, led by customer support (49%) and operations (47%)^[10]

Strategic Planning Assumptions:

• By 2027, 60% of teams will have a shared agent accessible via team chat
• By 2027, Slack's first-party Slackbot teammate will force at least two third-party platforms to reposition around data control or vertical depth
• By 2028, open-source self-hosted platforms will capture 30% of the market as teams demand auditability and data control

---

Market Definition

Team agent platforms deploy shared AI agents accessible across teams. Unlike personal agents (managing your individual email and calendar) or local IDE tools (your personal Cursor or Copilot), these agents are:

• Shared — Accessible by multiple team members across functions
• Asynchronous — Work continues without user presence
• Integrated — Live in Slack, Discord, or existing collaboration tools
• Action-oriented — Draft documents, analyze data, execute workflows, coordinate operations

Inclusion Criteria:

• Deploys agents accessible by multiple team members (chat, web workspace, or API)
• General-purpose team work (not solely software development)
• Can take action (documents, data, operations, workflows)
• Operates asynchronously (not just real-time chat)

Exclusion Criteria:

• Software factories and coding agent platforms (Devin, Tembo, Codex in Slack, Kilo, eksec) — primarily development-focused; covered in the coding-agent reports
• Agent-to-tool integration infrastructure (Composio, Nango) — covered in MCP Integration Platforms
• Personal AI assistants (Lindy, Manus, OpenClaw) — individual-focused; Lindy re-evaluated June 2026 and remains personal-assistant-first
• Pure frameworks (LangChain, AutoGen) — require significant integration
• Workflow automation converging on the category (Zapier Agents, Gumloop) — automation-builder identity; watchlist
• Mega-vendor ecosystems (Microsoft Copilot Agent Store, Gemini Enterprise) — OS-scale platforms, not comparable rows
• Chatbots without action capability

---

Adjacent: Enterprise-Managed Personal Agents

Grok Bot matters to the same buyers, but its organizing unit is a person. Each user gets a persistent cloud computer; all of that user's Bots share its files and signed-in applications. A different human receives a separate computer. Those are materially different boundaries from multiple people operating one shared agent and review queue.^[11][12]

Sharing a Bot distributes a template and gives the recipient their own copy. It does not transfer the original conversations, computer files, or logged-in accounts. That is useful for distributing a repeatable workflow across a company, but it does not establish shared runtime state or collective ownership of the original run. Grok Bot therefore remains in the personal-agent comparison, outside this report's 14-member matrix.^[13]

There are real organization controls: Teams includes access and selected policies, while Enterprise adds controls such as network policy, enforced Auto Review, and organization computer management. Do not infer those Enterprise controls from a Teams seat, or infer a shared agent from an organization-wide setting.^[11]

For a concrete comparison, consider a recurring customer-ticket roundup. A personal Bot can prepare an individual's review material; distributing its template lets coworkers run their own copies. Tembo Agent Studio instead documents a shared Tasks Inbox where a reviewer can edit a proposed action and execute it through their connected account. Evaluate who owns the agent, who can review the work, and whose credentials perform the action. Agent Studio's Git-based definitions and version promotion are additional decisions, not features established by Bot template sharing.^[13][7][14]

Disclosure: Ry Walker is Tembo's co-founder and CEO. This comparison describes distinct operating models; it does not claim a Grok Bot–Agent Studio integration or that every Agent Studio tool call requires review.

---

Comparison Matrix

---

Capabilities Matrix

Team agents vary widely in what they can do. This matrix rates each platform across six capability areas:

✅ = Strong | — = Limited/None

Agent Studio's row describes documented workflow mechanisms rather than assigning benchmark-derived strength scores. Its capabilities depend on the agent definition, tools, connections, and review policy.

Capability Definitions:

• Product Planning — PRDs, specs, requirements gathering, roadmap collaboration
• Coding — Write code, create PRs, fix bugs, refactor, code review
• Data Analysis — Query databases, generate reports, analyze metrics
• Customer Support — Answer tickets, route requests, knowledge base queries
• Sales/CRM — Pipeline updates, lead research, account management
• Operations — Workflow automation, scheduling, cross-team coordination

Key Insight: Dust and Glean now match the cross-functional breadth that previously only Claude in Slack offered — by grounding agents in company knowledge rather than chat context. Operations remains the capability nearly every platform covers.

---

Platform Profiles

Agentforce

Salesforce's enterprise AI agents — now Agentforce 360, with Slack as the front door^[16]

• Agentforce 360 — October 2025 rebrand unifying the agent platform, Data 360, and Slack; ~$800M ARR as of Q4 FY2026
• Multi-model — Spring '26 added Claude and GPT model choice
• Consumption pricing — Flex Credits ($500/100K; a standard action ≈ $0.10) plus per-user editions; budget unpredictability is the top complaint
• Salesforce-gated — Requires Salesforce (and usually Data Cloud) investment

Best For: Salesforce-centric enterprises needing business workflow automation.

---

Agent Studio

Tembo's MIT-licensed control plane for shared agents, with definitions in a GitHub repository and a separate runtime review surface. It is a distinct project from the commercial Tembo coding-agent platform. Hand-authored agents can run without the commercial service; adding a Tembo API key enables natural-language authoring and improvement requests.^[15] Disclosure: Ry Walker is Tembo's founder and CEO.

• Choose the change policy. Authoring defaults to reviewable PRs. A workspace admin can enable YOLO direct commits, subject to the repository's branch protections. Git history alone does not force a human approval.^[6]
• Separate draft from stable. The draft follows the repository file; promotion creates a numbered snapshot. Automated paths default to stable after the first promotion, while schedules can explicitly select draft. Owners and workspace admins control promotion.^[14]
• Operate shared work. Run agents manually, on schedules, or through events, webhooks, and Slack apps. The self-hosted stack includes the web app, Rust API, and Postgres.^[15]
• Review the proposed action. Tasks Inbox can stage a suggested reply or CRM/ticket action. The reviewer edits or selects the action, which uses their connected account; this is a configured workflow, not a guarantee that every agent tool call waits for approval.^[7]
• Inspect the external dependencies. Composio, native MCP providers, model APIs, and optional Tembo authoring have their own access and cost boundaries. Runs act through the relevant user's connections; hosting Studio yourself does not make those services local.^[8][15]

A useful pilot is a recurring ticket roundup that stages a few items for review, followed by a proposed change to its selection rules. Inspect the source action, the resulting Git diff, and the version used by the next scheduled run. This is an evaluation scenario, not a workload benchmark performed for this report.

Best For: Teams that want to operate shared agents from versioned specifications and can maintain the deployment, credentials, and review policy. See the full Agent Studio report for deployment, licensing, evidence limits, and the distinction between current repository behavior and pinned releases.

---

AgentOS (Agno)

Self-hosted agent runtime from the highest-starred framework — your VPC, your data^[17]

• Runs in your infrastructure — FastAPI runtime + browser control plane; agent data never leaves your VPC
• Multi-surface — Slack, Discord, Telegram, WhatsApp, and API
• Model-agnostic — 30+ providers on Agno's 40K-star Apache 2.0 framework
• Developer-deployed — A runtime you build a platform on, not a turnkey SaaS workspace

Best For: Engineering-led teams that want shared agents with full data control and chat surfaces.

---

Claude in Slack

Anthropic's native Slack integration — broadest capabilities from coding to product planning^[4]

• Cowork mode — Product planning, research, writing, and cross-functional collaboration
• Claude Code mode — Full coding capabilities with PR creation and code review
• Thread context — Gathers context from entire Slack threads
• Team visibility — Progress visible to entire channel; seamless web handoff
• Claude-only — No multi-model support; $20–125/seat/mo depending on tier

Best For: Teams wanting the broadest capability set — strong across planning, coding, data, and support.

---

CrewAI Enterprise

Multi-agent orchestration with visual editor — 63% Fortune 500 adoption^[18]

• Multi-agent workflows — Agents collaborate with role-based specialization
• Visual editor — Build crews without code; free tier now includes tracing
• Open source core — Apache 2.0 licensed; 53K+ GitHub stars
• Simplified pricing — Free tier + custom Enterprise (FedRAMP High, dedicated VPC); the public mid-tier is gone

Best For: Teams wanting multi-agent patterns with open source foundation.

---

Dust

"Multiplayer AI" — 3,000+ orgs running 300K+ shared agents on company data^[2]

• Shared agent workspace — Teams build and share custom agents grounded in company knowledge
• Multi-model — GPT, Claude, Gemini, Mistral, user-selectable per agent
• Slack-native plus web — Agents callable in channels; Zendesk, Chrome extension, API
• Commercial proof — 240% net revenue retention, zero churn in 2025; $40M Series B (Abstract + Sequoia, May 2026)
• Open core — MIT-licensed repo; the hosted SaaS is the product

Best For: Teams that want many purpose-built shared agents over company data, fast.

---

Glean

Enterprise Work AI at $7.2B — shared agents over permissions-aware company knowledge^[3]

• Agent Library — No-code builder plus shared library, deployed over 27B+ indexed documents
• Permissions-aware — Agents respect existing document-level access controls
• Multi-model Model Hub — OpenAI, Anthropic, Google, Amazon, Meta per agent
• Enterprise-only — Sales-quoted pricing (est. $35–50/user/mo); agents are one pillar of a broader search/assistant suite

Best For: Large enterprises wanting governed agents over all company knowledge.

---

MindStudio

No-code visual agent builder — $36M raised, 200+ models behind one router

• Visual workflow builder — agents deploy as web apps, scheduled jobs, browser extensions, email agents, webhooks, or MCP servers
• Multi-model — 200+ models via its Service Router or BYOK, pass-through pricing
• No chat-native surface — no Slack/Teams deployment; web, API, and extensions instead
• Reclassified here in June 2026 — previously listed with agent-company platforms, but it's a workflow builder, not an org-structure platform

Best For: Non-technical operators and agencies building scheduled, autonomous agent workflows.

---

Runbear

No-code AI agent platform — 300K+ users, autonomous triggers since May 2026^[19]

• No-code builder — Prebuilt agents plus AI-assisted agent builder
• Multi-platform — Slack, Teams, Discord, HubSpot, Zendesk
• Autonomous now — Scheduled and event-driven triggers shipped May 2026
• Flat pricing — Team $79/mo, Business $319/mo (BYOK); credit-based, no per-seat fees

Best For: Ops teams wanting no-code agents across multiple platforms.

---

Sim AI

YC-backed open-source visual agent builder — now "The AI Workspace" with Mothership^[20]

• Figma-like canvas — Drag-and-drop workflow editor; no boilerplate
• Mothership — Natural-language control plane orchestrating workflows, tables, files, and integrations (March 2026)
• Multi-LLM — OpenAI, Claude, Gemini, Mistral, xAI, and local models; 1,000+ integrations
• Self-hostable — Apache 2.0; 28.8K stars as of June 2026; $7M Series A

Best For: Teams wanting a visual, open-source workflow builder that non-engineers can use.

---

Spacebot

Spacedrive's Rust agent framework for teams and communities — concurrent by design^[21]

• Five process types — Channel, Branch, Worker, Compactor, Cortex split the monolith into specialized concurrent processes
• Built for communities — Handles 50+ concurrent users without blocking on any single conversation; Discord-native
• Credible team — Jamie Pine's Spacedrive (35K+ stars) pedigree; same Rust and distributed-systems DNA
• Source-available — FSL license converts to Apache 2.0 after two years; alpha maturity

Best For: Communities and teams wanting a self-hosted, multi-user agent that stays responsive under concurrent load.

---

Tasklet

Plain-English cloud agents from the Firebase co-founder — now team-focused with $20M raised^[22]

• No workflow builder — Describe in English; AI figures out implementation
• Fresh capital — $20M (2026) at a $175M valuation; $5M ARR
• Flexible triggers — Scheduled, email, webhooks; always-on cloud agents
• One-way Slack — Agents post reports into channels; conversational control lives in the web app

Best For: Business process automation; not specifically for coding workflows.

---

Tensol

Managed OpenClaw for B2B — VM isolation, OAuth, audit logs (YC W26)^[23]

• 5-minute setup — No infrastructure to manage
• Per-customer isolation — VM-level separation for enterprise
• Web-presence churn — Site moved to tensolai.com; pricing still undisclosed
• Early stage — 2-person team; upstream OpenClaw shipping its own enterprise features is a squeeze risk

Best For: Teams wanting managed OpenClaw with enterprise security — eyes open on vendor maturity.

---

Viktor

The AI coworker with its own computer — $75M Series A, $15M run-rate in three months^[1][24]

• Own compute — Persistent cloud environment for code execution and app building
• Slack and Teams — Embedded as a coworker in both; 20,000+ workspaces as of June 2026
• 3,200+ integrations — Browser automation + native APIs
• Backed by Slack's founders — Accel-led round with Butterfield and Henderson as angels

Best For: Teams wanting a chat-native agent with autonomous compute and the strongest momentum in the category.

---

Architecture Patterns

First-Party vs. Builder vs. Managed

Slack-First vs. Multi-Channel

---

Gap Analysis

Key gaps across the market:

• Discord support is still rare — only AgentOS and Runbear; Discord-first teams remain underserved
• Connected context needs its own evaluation — Agent Studio's authorized tools can reach company systems; that does not establish the same indexing, retrieval, or permission behavior as a company-search product. Test the required sources and acting identity.^[8]
• Version history is not the whole review policy — Agent Studio separates Git changes, stable-version promotion, and Tasks Inbox decisions. Verify which controls are enabled rather than assuming that every edit or external action requires approval.^[6][14][7]
• First-party channel risk is unpriced — Slackbot-as-teammate threatens the thin-wrapper end of the category most

---

Strategic Recommendations

By Use Case

By Buyer Profile

Startups (< 20 people): → Claude in Slack if you're already on Claude. Dust or Sim AI for shared agents non-engineers can build.

Scale-ups (20-200 people): → Dust for knowledge-grounded shared agents; Viktor for an agent with compute; Agent Studio or AgentOS if engineering wants agents under version control or in-VPC.

Enterprise (200+ people): → Glean for governed agents over indexed knowledge; Agentforce if Salesforce-centric; Claude in Slack for standardized tool stacks.

Open Source-First Teams: → Sim AI (Apache 2.0, 28.8K stars) is the most mature open option; AgentOS for chat-surface runtimes; Agent Studio (MIT) for git-native rigor.

---

Market Outlook

Near-Term (2026)

• The funding wave continues — expect at least one more $50M+ round and the first acquisition of a mid-tier player
• Slack's first-party Slackbot teammate ships broadly, pressuring thin Slack-wrapper products^[9]
• Open-source platforms (Sim AI, Agent Studio, AgentOS) compete on self-hosted auditability as agent governance becomes a procurement requirement

Medium-Term (2027-2028)

• 60% of teams will have a shared agent accessible via team chat
• Knowledge grounding becomes table stakes — expect Dust/Glean-style company-context features in every platform
• Versioned, reviewable agent configuration becomes a standard enterprise requirement

Long-Term (2029+)

• Team agents will be default infrastructure, like CI/CD
• The distinction between "workflow automation" and "team agent" will blur completely
• Agent governance (who changed what agent, when, and why) will be a compliance topic

---

Bottom Line

Team agent platforms graduated from experimental to venture-validated in the first half of 2026. The market now has four distinct camps:

Current Leaders:

• Claude in Slack for the broadest capability set on a single backend
• Glean for enterprise-scale agents over governed company knowledge
• Dust for multiplayer shared agents with the strongest commercial proof points

Rising Challengers:

• Viktor — the fastest-growing, with compute-equipped agents and Slack's own founders backing it
• Sim AI, Agent Studio, AgentOS — the open-source/self-hosted wing, betting governance and data control win the enterprise

The key strategic question: Do you want agents grounded in your company's knowledge (Dust/Glean), embedded in your chat with maximum capability (Claude), or under your own version control and infrastructure (Agent Studio/AgentOS/Sim AI)?

For most teams, the answer will evolve. Start with first-party or knowledge-grounded platforms for speed. Move toward open, self-hosted options as governance, auditability, and multi-model needs grow — and watch what Slack does with Slackbot.

---

Research by Ry Walker Research • methodology

Disclosure: Author is CEO of Tembo, which makes Agent Studio.