Alternatives to Kaiku

Tools: enabled with custom skills, tools, and APIs

Does: Triage my inbox Update the CRM Draft a reply Summarize Slack

For: Financial Services, Manufacturing, Pharma & Biosciences, Technology & Software, Higher Education, Human Resources, IT, Sales, Marketing, Finance, E…

Does: Answers prospect and customer questions built from the customer's own product information and company knowledge. The cited span does not support "t…

Tools: Uses tools via integrated marketplace with dynamic routing and RAG support

Appsmith Agents is closed to new users. The page these fields were extracted from - agents.appsmith.com, captured 2 August 2026 - states that "Appsmith Agents is no longer accepting new sign-ups" and that "Existing users keep access while we integrate agent capabilities into an upcoming update." The same page labels the product "Appsmith Agents (Legacy)" and offers a waitlist rather than signup. The evidenced fields below describe what the product offered to existing users at the time of that capture: deployment on-prem, in a VPC or cloud provider, or on managed hosting; agents embedded in Gmail, Salesforce, Zendesk and Slack; role-based access controls extending to agent data and functions; audit logs and monitoring dashboards; and responses carrying automatic citations from customer data. They should not be read as describing a product a new customer can buy today. AI Dispatch has held no capture of this page newer than 2 August 2026, so the current state of the integration referred to in the notice is unknown.

For: Voice Agents, AI Notetakers, Call Analytics, Medical Transcription, Dictation, Agent Assist, AI Scribes

Data is not used for training; retained only for operational purposes within customer workspace; deletable via revocable access

Data: upholds the highest security, compliance, and sovereignty standards

Birta is a publishing and hosting service for sites built by a user's own AI chat agent: the agent builds the site, Birta puts it at an address of its own reachable over a secure connection, and keeps every earlier version in a panel, each saved with its date and the note the agent left for it. Viewers can comment on a page without an account, and everyone added to a project gets identical rights - there are no tiered roles. Documented setup paths cover Claude, ChatGPT, Claude Code and Cursor. Its homepage displays a "Verified by Anthropic" badge and describes itself as a verified connector in Claude's directory; both are Birta's own claims and we have not confirmed either with Anthropic. Pricing is freemium, from a free tier to a $39/month "Pro" tier. Site files can be downloaded at any version and a connected custom domain remains the user's own. It is explicitly not intended for web apps with sign-ups and logged-in areas.

Does: Autonomously publishes content directly to the customer's CMS

Commt is an AI agent platform on which each agent is bound to the knowledge bases and tools the customer connects to it, with retrieval-augmented generation optional. Agents can run on frontier models from Anthropic, OpenAI and Google or on open-source models such as Llama and Qwen, and the vendor states switching model is a dropdown change that carries instructions and knowledge bases over unchanged. Twenty built-in connectors are listed, spanning messaging and email, work and knowledge tools, and customer and sales tools; agents can be extended with custom MCP servers, which the vendor says run in an isolated environment and are checked for security issues before use. Application, database and document storage run in the EU on private networking, and Enterprise engagements are offered private-server or on-premises deployment with an uptime SLA and named support. On controls, the vendor states that each agent's access to knowledge bases, tools and the internet is set by the customer, that the monthly request limit is checked before each request so a runaway loop cannot consume the remaining allowance, that customers configure refusal topics and low-confidence behaviour, and that every answer cites its sources. By default no message text is written to storage; what is retained is usage metadata such as request counts, latency and which agent answered. Pricing is fixed monthly plans with a defined request allowance rather than metered billing, with the entry plan listed at $244 a month from $349 under a 30% early-access discount. The product was not generally available when captured: it was in an early-access round of 25 seats, with the vendor stating doors open on 12 October 2026. AI Dispatch found no public information, in the source captured, on any compliance certification or attestation (no SOC 2, ISO 27001 or HIPAA claim appears), on any stated EU AI Act or NIST AI RMF posture, on insurance or customer indemnity, or on a tamper-evident audit trail -- the vendor's own statement that message text is not stored by default is inconsistent with a reviewable record of what an agent said. Every field on this entity is evidenced from a single capture of the vendor's own homepage taken on 2026-09-27; no independent or third-party source corroborates these claims, all of which are the vendor's own.

Cortx is an enterprise AI agent platform from XNODE Inc. Per the vendor's own site, it runs agents that retrieve data, execute tasks and complete workflows end to end: a goal is broken down, acted on, and the agent checks in with the customer's team before consequential decisions, with maker-checker approval gates a human reviewer must clear before downstream actions proceed. Guardrails are described as policies on what AI accesses, generates and shares, enforced automatically rather than reviewed after the fact, alongside role-based access control. The platform is stated to be Generally Available and live in production, deployed on premises, AKS and EKS within the customer's network boundary, and MCP and A2A native. The vendor states Cortx operates across hedge funds, asset managers, diversified financial conglomerates and government agencies in the US and India, and that it is certified to ISO 27001, ISO 42001, GDPR, HIPAA and SOC 2 Type II - self-declared, with no certificate, auditor or report date published on the page checked. Customers may use private LLMs or connect custom models; the platform is stated to use third-party AI services, which it does not name. A full audit trail is claimed, with no tamper-evidence or hash-chaining. Pricing scales with deployment size and usage. No insurance, indemnity or outcome-linked payment term is published; the terms instead disclaim liability for any direct, indirect, incidental, consequential or punitive damages to the fullest extent permitted by law.

Does: Browser control (navigate/fill/click/read) plus messaging bridges, tool/skill execution and database queries

Credal builds governed Model Context Protocol (MCP) servers for enterprise teams, with permissioning, a consolidated audit trail, and connections to systems like Google Drive, Slack, Salesforce, and Snowflake.

For: sales, marketing, operations, and finance teams.

Dify is a platform for building agentic workflows and RAG pipelines. Per the vendor's own homepage, agents built on it reason through a task, use approved tools, keep context and stop within clear limits, and finished apps can be published as hosted experiences, API endpoints, embeds or MCP-compatible tools. Model providers, tools, data sources and MCP integrations are installed from a marketplace or defined by the customer. It is offered as hosted SaaS and can also be deployed to a VPC or self-hosted; the vendor describes an open-source community edition that is source-available under an Apache-2.0-derivative licence, self-hostable with Docker, and reports 151K+ GitHub stars. On assurance, the page names SSO/SAML, RBAC and audit logs, and states SOC 2 Type II and ISO 27001 - self-declared, with no certificate, auditor or report date published on the page checked. Plans are described as scaling with team, storage, logs and usage, with a dedicated CSM and 24/7 support on higher tiers. No insurance, indemnity, coverage limit or outcome-linked payment term is published on the page checked.

For: Legal ops, finance, or procurement leads at 50-to-300-person companies with a growing pile of vendor and SaaS contracts

Does: Agent can act via web browsing, API calls, document handling, and voice calls, beyond just chat replies.

Fabrile is a customer-support agent that answers from a business's own documents, pages and help articles and links each answer back to its source. The vendor states it "only speaks from the docs, pages, and help articles you give it - not the open internet" and that it escalates to a human team, with replies handled from a single inbox. It embeds with one line of code and offers one-click installation for Squarespace, Shopify and Webflow. On data handling the vendor says it never trains on customer content or resells it and that conversations run on EU infrastructure. Compliance claims are self-declared: "ISO 42001 aligned" and "EU AI Act ready", with no certificate, auditor, scope or audit date published on the page checked. Pricing starts free, with full refunds within 30 days and pro-rating thereafter. No audit trail of agent activity, no third-party evaluation, no insurance, indemnity or outcome-linked payment term is published on the page checked.

For: marketing, SEO, sales, and customer support

For: financial services, healthcare, and travel sectors

For: Tech, Financial Services, Insurance, Retail, Automotive, Education, Healthcare, Real Estate, Telecom, Professional Services, Venture Capital

Tools: Agents select the wrong tool or call the right tool the wrong way

Data: Retains personal data only as long as necessary for its purposes and legal compliance; no fixed period stated

HarnessRouter sells embedded infrastructure for running third-party agent harnesses - Codex, Claude Code, Hermes and others - in isolated per-task sandboxes behind a single API, so a customer product can send a task and get back finished work such as videos, games, codebases or documents. The harness and model are named in config and swappable with a one-line change, and requests go to whichever model provider and credentials the customer configures. It is offered both as a managed Cloud service and as a self-hosted Apache 2.0 Community Edition in which runtime and application state stay in the customer environment. Agent work is metered from $0.014 per active work minute with waiting for model responses unbilled, and plans start at $20 per month applied as usage. The company publishes Y Combinator backing and names Stanford Medicine as a launch partner. On the assurance questions this index scores, the published record is thin. Every session runs in its own isolated sandbox, which bounds blast radius, but no scope, spend or authority limit on the agent itself is published. The only record of agent activity is an execution trace streamed to the calling application, with no stated retention period and no tamper-evidence. No certification, carrier, cover limit or customer indemnity is named anywhere in the capture. Currency note - this page was substantially rewritten between the 2026-09-04 and 2026-09-11 captures, which added the Apache 2.0 Community Edition and dropped the per-minute rate. Eight fields, the two pricing fields among them, still cite the earlier capture and are marked as dated where the newer capture no longer carries the wording. Re-extraction against the newest capture is pending with the Researcher.

Does: Provides a 'Matrix' interface letting users run large-scale research queries (add rows/columns) across many source documents at once, rather than a…

Does: navigating pages, filling forms, and extracting data

Tools: integrates with other tools (like Slack, Gmail, or Salesforce)

Does: connect tools

Integrates: OpenAI-compatible API; the vendor states its endpoint is fully OpenAI compatible.

Does: Can complete a purchase/booking using a single-use virtual card capped to the approved amount

Audit trail: Vendor states every answer traces back to approved knowledge with a full audit trail

For: healthcare, HR, public administration, education, technology, and marketing

Governance: Jump in while it runs: point at any element on the page, send a note, or change direction and it folds your input in without starting over. Rewind …

Explainability: Walk through each step of your AI agent’s reasoning, from the raw input it receives to the decision it returns.

Tidio Copilot is a customer-support assistant delivered as a free Chrome extension that suggests replies to support agents based on conversations handled across live chat, email and social media. The vendor states it is built on Anthropic's Claude AI and that it draws on a custom knowledge base built from the customer's website content and FAQs. It is published as working with Zendesk, Intercom, Gorgias and Gmail, and as completely free to use. The vendor states the product is GDPR and CCPA compliant and that data is encrypted at rest and in transit; both are self-declared, with no certification, auditor or report date published on the page checked. No adoption figure is published for Copilot itself - the 300,000+ businesses figure on the page describes the wider Tidio platform Copilot is built into. No audit trail, run-time limit, independent evaluation, insurance, indemnity or outcome-linked payment term is published on the page checked.

Audit trail: Audit trail stated as a bundled plan feature with the vault

For: Human resources, Healthcare, Banking and financial institutions, Supply chain, Insurance, Public sector

Audit trail: Vendor states each added department inherits the same audit trail and controls

Integrates: OpenAI, Spotify, YouTube, Twitch, Steam, GitHub, Notion, Reddit, +9 more

Does: Agents carry long, multi-step work across many systems (ticketing, data platforms, cloud accounts, internal apps) through to completion, with budge…