Telnyx vs Twilio (2026)

Assistants call outside systems with webhook tools, which can run asynchronously with filler messages while the request completes, attach MCP servers for whole tool catalogs, and connect enterprise integrations for Salesforce, Zendesk and Jira. Call control tools transfer the caller, hand off with SIP Refer, send DTMF tones to another phone system and hang up, client side tools run actions in the web widget, and keys for outside providers are kept as write only integration secrets. Tools live in a shared library that several assistants and workflow nodes can reuse.

Sourcedevelopers.telnyx.com/docs/inference/ai-assistants/no-code-voice-assistant.mdread 2026-10-11

Agent Connect gives an agent a tool system built on standard function calling. A developer defines a name, description and JSON schema, and the SDK formats the tool for whichever LLM provider runs the agent. Custom tools are written with a decorator in Python or defineTool in TypeScript and can read the session, the customer profile and the channel.

Three tools come built in, knowledge search over Enterprise Knowledge, memory retrieval and escalation to a Studio flow that can route into Flex. Because each tool is the customer's own code, the agent can act in any system the customer can reach, and Twilio's own messaging, voice, email and Verify APIs sit alongside as further actions.

Sourcetwilio.com/docs/conversations/agent-connect/core-conceptsread 2026-10-11

Conversation workflows store a directed graph on the assistant. Prompt nodes run the model with their own instructions and optional model and voice overrides, speak nodes play a fixed message, and tool nodes run one shared tool and advance.

Edges move between nodes on a natural language condition the model judges, a deterministic comparison against variables, or a default, and a tool node's HTTP status code can steer the next edge.

An edge can also hand the conversation to another assistant, so a triage assistant can route billing, technical and sales callers to specialists that share the conversation's context.

Sourcedevelopers.telnyx.com/docs/inference/ai-assistants/workflows.mdread 2026-10-11

Conversation Orchestrator joins voice calls, SMS and WhatsApp messages into one threaded conversation tied to a customer profile, with a lifecycle that moves each conversation from open to inactive to closed on timeouts the customer sets. Studio, Twilio's visual flow builder, runs multistep communication workflows and receives the agent's handoff payload, and Flex routes the conversation to a person with the history attached. The reasoning inside the agent stays in the customer's own code, while the movement of a conversation between channels, the agent, flows and people runs on Twilio.

Sourcetwilio.com/docs/airead 2026-10-11

Calls reach an assistant on an assigned Telnyx number or over SIP, outbound calls start from the portal, the API or the TeXML ai_calls endpoint, and an embeddable widget carries voice and chat on a web page. SMS conversations run through the assistant SMS chat API, with MMS during a call when messaging is on. At the start of each conversation a dynamic variables webhook can load caller details and memory settings, and custom SIP headers or the outbound request can pass variables in.

Sourcedevelopers.telnyx.com/docs/inference/ai-assistants/no-code-voice-assistant.mdread 2026-10-11

Inbound voice calls, SMS, WhatsApp, RCS and web chat all reach the agent through one shared callback, so a single handler serves every channel. Webhooks and WebSocket connections carry the events, and Conversation Relay streams a live phone call to the agent with speech recognition and synthesis handled by Twilio. Outbound sends and calls start from the REST APIs, and Studio flows start on incoming calls, messages or API requests, which covers inbound, event driven and programmatic starts across the channels a contact center uses.

Sourcetwilio.com/docs/conversations/agent-connect/core-conceptsread 2026-10-11

Each assistant can use named knowledge bases built from uploaded files or a URL, kept with the account and attached in the assistant's configuration, so one knowledge base can serve several assistants. Retrieval during a call is part of the voice engine's orchestration and included in the per minute price, while the price estimator carries a separate agent knowledge base add on. Telnyx does not say how the content is chunked or indexed.

Sourcetelnyx.com/pricing/conversational-airead 2026-10-11

Enterprise Knowledge stores a company's policies, documents and reference material as knowledge bases that agents and human agents search for grounded answers. It is billed as a persistent store, $0.018 per GB hour of storage and $0.005 per retrieval, and Agent Connect exposes it as a knowledge search tool the agent calls when a customer asks a question, returning matching passages for the reply. Customer profiles live separately in Conversation Memory.

Sourcetwilio.com/en-us/products/conversational-ai/pricingread 2026-10-11

Memory brings earlier conversations into a new one. The dynamic variables webhook returns a query in the same form as the List Conversations endpoint, typically the caller's phone number, a time window or custom metadata, plus the IDs of insights to include, and the assistant starts with those conversations and insight results in context, across voice and messaging.

The webhook has one second to answer before the call carries on without memory. What comes back is stored conversation history picked by the customer's query. There is no separate memory store, and Telnyx sets out no retention period or way to delete one person's history.

Sourcedevelopers.telnyx.com/docs/inference/ai-assistants/memoryread 2026-10-11

Conversation Memory keeps one profile per customer inside a memory store. Each profile holds conversational memory, the observations and summaries extracted automatically after each conversation closes, and factual memory, the traits and identifiers that describe the customer. New observations are added, merged with an existing one when they add detail, replace it when they contradict it, or are discarded when they lack support, so the profile stays current.

A Recall API returns the most relevant observations, summaries and recent messages by semantic and lexical search, and Agent Connect calls it on every messaging turn. Up to five stores separate data such as sandbox and production, each profile holds up to 500 observations and 500 summaries, and observations can be switched off per configuration. Memories carry no fixed retention period and persist until they are reconciled, edited or removed.

Sourcetwilio.com/docs/conversations/memory/conversation-memory-conceptsread 2026-10-11

Transfer and SIP Refer tools pass a caller to a person or another phone system, with voicemail detection on transfer, and handoff tools move a conversation between assistants. Releases stay under the team's control. A new version is saved with notes, traffic rules send selected callers or a percentage to it, and rollback returns all traffic to the main version at once. No step lets a person approve a tool call or a reply before the assistant acts.

Sourcedevelopers.telnyx.com/docs/inference/ai-assistants/version-testing-traffic-distribution.mdread 2026-10-11

The built in escalation tool hands a conversation to a Studio flow and on to a Flex agent with full context, on voice after the agent's last reply plays and on digital channels by posting straight to the flow. When and why to escalate is the customer's code to decide, and Agent Connect makes no decisions about what the agent does.

Conversation Memory applies a prompt exclusion meant to screen out GDPR special category data. Agent Identity adds scoped, revocable tokens and human approval over SMS or RCS before a high risk action, but it is in private release, so approval before each action is not yet open to every customer.

Sourcetwilio.com/docs/conversations/agent-connect/core-conceptsread 2026-10-11

Business accounts sign in to the Mission Control Portal through SAML single sign on with Okta, Azure Active Directory, Google, OneLogin, Auth0 or LastPass. Telnyx holds ISO 27001:2022 and ISO 27701, a SOC 2 Type II report under NDA covering voice, messaging, video and wireless, and HIPAA coverage for voice, messaging, fax and AI inference with a BAA available, plus PCI DSS for its own billing. The SOC 2 scope does not name AI assistants, and Telnyx also claims a SOC 2 Type III, a report type the AICPA does not issue.

Sourcetelnyx.com/ai/compliance.mdread 2026-10-11

Twilio Console sign in works through a corporate identity provider over SAML 2.0, with SP and IdP initiated login and multiple SSO profiles, on Twilio Editions that include single sign on. Which accounts a user can open and which role they hold are set in the Console and Admin Center, and Flex maps identity provider roles to its own.

Twilio holds ISO/IEC 27001 certification and is audited externally on a regular cycle, with security and privacy reports available on request through security.twilio.com. Agent Connect and Conversation Memory are not HIPAA eligible or PCI compliant, so regulated workflows sit outside them.

Sourcetwilio.com/docs/iam/single-sign-onread 2026-10-11

Every conversation is stored and listable through the conversations API, transcripts show which workflow node produced each assistant message, and the last tool call's HTTP status is kept as a variable. Insight groups run structured and unstructured analysis on each conversation and can post results to a webhook, with Telnyx managed insights alongside custom ones. Version notes record what changed in each release, and Telnyx does not mention an audit trail of who changed an assistant's configuration.

Sourcedevelopers.telnyx.com/docs/inference/ai-assistants/workflows.mdread 2026-10-11

Conversation Orchestrator stores every message and transcribed call as a threaded conversation linked to a profile, Conversation Memory shows the observations and summaries it extracted, and Conversation Intelligence analyzes each conversation for sentiment, summaries and custom criteria. Agent Connect logs a warning when it falls back from the Memory API. Together that shows what was said and what the system learned. The tool calls and reasoning live in the customer's agent, so a step by step trace of a run comes from the customer's own logging.

Sourcetwilio.com/docs/conversations/memoryread 2026-10-11

Voice AI agents run in the United States, the primary data center, or in the European Union with EU data residency, where data does not leave the EU. AI inference is also offered in Asia Pacific with regional processing, though voice agents are not offered there, and Canada has neither. The assistant's language model can also point at the customer's own OpenAI compatible endpoint, such as one on AWS Bedrock, Azure OpenAI or a self run vLLM server, which keeps inference inside an environment the customer controls.

Sourcetelnyx.com/ai/compliance.mdread 2026-10-11

The agent and its model run wherever the customer hosts them, since Agent Connect is an SDK inside the customer's own application. Twilio's isolated regions are US1 in the eastern United States, IE1 in Dublin and AU1 in Sydney, and a customer chooses one to keep a workload's data there.

Voice, SIP trunking and messaging run in Dublin, and voice in Sydney, while Conversation Memory, Orchestrator, Enterprise Knowledge and Conversation Relay are offered in the US region only, so the agent layer runs in the US. While regions roll out, Twilio does not guarantee that all data stays in the selected region.

Sourcetwilio.com/docs/global-infrastructure/regional-product-and-feature-availabilityread 2026-10-11

New assistants start from a template in the portal, with a blank template as the usual starting point, and Telnyx managed insights come ready to attach to any assistant. A shared tools library lets one set of tools serve many assistants. Telnyx offers no catalog of finished agents for support, sales or scheduling.

Sourcedevelopers.telnyx.com/docs/inference/ai-assistants/no-code-voice-assistant.mdread 2026-10-11

Agent Connect starts from a quickstart and integration guides for AWS Bedrock and Microsoft Foundry, Conversation Intelligence ships Twilio authored language operators alongside custom ones, and Twilio Skills and the Code Exchange give developers guided patterns and sample apps. Each of these is a starting point the customer builds on. Twilio does not offer finished agents for support, sales or scheduling that a customer switches on.

Sourcetwilio.com/docs/conversations/agent-connectread 2026-10-11

The customer picks the model for each assistant on the Agent tab. Options include models hosted on Telnyx GPUs, with Kimi as the default, managed third party models from providers such as OpenAI, Gemini and Groq, with OpenAI models on the customer's own key, and any public OpenAI compatible endpoint such as AWS Bedrock, Azure OpenAI, Baseten, vLLM or SGLang. Prompt nodes in a workflow can override the model step by step, and speech to text and text to speech providers are chosen separately.

Sourcedevelopers.telnyx.com/docs/inference/ai-assistants/custom-llm.mdread 2026-10-11

The customer picks the model. Agent Connect does not run inference. The customer supplies the LLM application and controls how it responds, with integration guides for AWS Bedrock and Microsoft Foundry and support for OpenAI or any other provider. Tools are formatted for whichever provider the customer uses, so switching models is a change in the customer's own code. Twilio's own generative features, such as observation extraction in Conversation Memory, run on models Twilio chooses.

Sourcetwilio.com/docs/conversations/agent-connectread 2026-10-11

The API covers assistants end to end, including creating and updating assistants, assistant tests, canary deploys, MCP server records, SMS chat and a chat endpoint in beta, and the TeXML ai_calls endpoint starts outbound calls. Coding assistants can read the whole reference through llms.txt indexes, and the conversations API hands stored conversations to other systems.

Sourcedevelopers.telnyx.com/development/llms/ai-assistants-llms-txtread 2026-10-11

Every product has a public REST API and helper libraries, including a Conversation Memory API with profile, trait, observation, summary, import and Recall endpoints. Agent Connect itself is an SDK for Python and TypeScript with a FastAPI server for webhooks and WebSockets. An MCP server gives coding agents semantic search across more than 1,500 endpoints in over 30 products, and Twilio Skills package product guidance for AI coding tools.

Sourcetwilio.com/docs/airead 2026-10-11

AI tests give each check a name, a target assistant and success criteria, such as what the greeting must say, and a test run shows progress live and reports whether the assistant met every criterion, with the conversation to review. A new version can be saved with notes and sent a share of traffic by ordered rules on the caller's number, all to one version or split by percentage, with the remainder staying on the main version, and rollback clears every rule at once. Coval can generate test conversations from seed cases.

Sourcedevelopers.telnyx.com/docs/inference/ai-assistants/version-testing-traffic-distribution.mdread 2026-10-11

Conversation Intelligence runs language operators over voice and messaging conversations, Twilio authored ones for sentiment and summaries and custom ones written as natural language prompts, which a team can point at its AI agent's conversations to flag compliance misses or missed upsells. That reviews live traffic after the fact. The Agent Connect quickstart builds and tests an application locally, and there is no suite of saved test cases with pass or fail results to run against a new version of an agent before it goes live.

Sourcetwilio.com/en-us/products/conversational-ai/pricingread 2026-10-11

Assistants act through webhooks, MCP servers, integrations and call control tools, and send keypad tones to other phone systems. No hosted browser, remote desktop or screen control is offered.

Sourcedevelopers.telnyx.com/docs/inference/ai-assistants/tools-library.mdread 2026-10-11

The agent layer connects an agent to communication channels, customer memory and knowledge through APIs and tools the customer writes. No hosted browser, remote desktop or interface control is part of it. An agent that has to work inside a web page or a desktop application brings that capability as a custom tool in the customer's own code, outside anything Twilio runs.

Sourcetwilio.com/docs/airead 2026-10-11