What problem do stateful MCP client capabilities solve on Amazon Bedrock AgentCore Runtime?

Stateful Model Context Protocol (MCP) client capabilities on Amazon Bedrock AgentCore Runtime address the critical limitations of previous stateless AI agent implementations. Stateless agents struggled with interactive, multi-turn workflows, as they couldn't pause mid-execution to solicit user input for clarification, request dynamic large language model (LLM)-generated content, or provide real-time progress updates during lengthy operations. Each request was independent, lacking shared context. This new feature fundamentally transforms agent interactions by enabling bidirectional conversations, allowing agents to maintain conversational threads, gather necessary input precisely when needed, generate dynamic content on the fly, and transparently inform users about ongoing processes. This leads to the development of significantly more responsive, intelligent, and user-centric AI applications capable of complex, adaptive workflows.

How does the transition from stateless to stateful mode work on AgentCore Runtime?

The transition to stateful mode within Amazon Bedrock AgentCore Runtime is initiated by a simple configuration adjustment: setting `stateless_http=False` when starting your MCP server. Once enabled, AgentCore Runtime provisions a dedicated microVM for each individual user session. This microVM is designed for persistence throughout the session's duration, which can last up to 8 hours or expire after 15 minutes of inactivity, ensuring isolated CPU, memory, and filesystem resources for each session. Continuity across interactions is maintained through a unique `Mcp-Session-Id` header. This ID is established during the initial handshake and subsequently included by the client in all follow-up requests, ensuring they are accurately routed back to the correct, persistent session, thereby preserving context and enabling complex, interactive dialogues.

What is Elicitation, and how does it enhance AI agent interactions?

Elicitation is a powerful stateful MCP capability that allows an AI agent (acting as the MCP server) to intelligently pause its ongoing execution and request specific, structured input directly from the user via the client. This significantly enhances interactive agent workflows by enabling agents to ask targeted questions at precise, opportune moments within their operational flow. For example, an agent might use elicitation to confirm a decision, gather user preferences, or collect particular data values that are contingent on preceding steps. The feature supports two robust modes: 'Form mode' for direct structured data collection through the MCP client, and 'URL mode' for secure, out-of-band interactions that require directing the user to an external URL (e.g., for OAuth or sensitive credential entry). The user's response – whether accepting, declining, or canceling the request – is then returned to the server, allowing the agent to dynamically adapt its workflow based on real-time human feedback.

How does Sampling capability benefit AI agents without managing LLM credentials?

Sampling equips the MCP server with the ability to request sophisticated large language model (LLM)-generated content directly from the client using the `sampling/createMessage` mechanism. A key benefit is that the MCP server itself does not need to manage its own LLM credentials, API keys, or direct integrations with various LLM providers. Instead, the server simply provides a well-formed prompt and any optional model preferences to the client. The client then acts as an intelligent intermediary, forwarding this request to its connected LLM and returning the generated response back to the server. This abstraction allows AI agents to seamlessly leverage powerful language model capabilities for tasks such as crafting personalized summaries, generating natural-language explanations from complex structured data, or producing context-aware recommendations, all while simplifying the operational overhead and security concerns associated with LLM management on the server side.

Amazon Bedrock: Uwezo wa Mteja wa MCP wenye Hali (Stateful) kwenye Wakati wa Kuendesha wa AgentCore

Kuboresha Mawakala wa AI: Mabadiliko kuelekea MCP yenye Hali (Stateful) kwenye Amazon Bedrock

Mawakala wa AI wanabadilika kwa kasi, lakini uwezo wao kamili mara nyingi umekuwa ukizuiliwa na utekelezaji usio na hali (stateless), hasa katika hali zinazohitaji mwingiliano wa mtumiaji wa wakati halisi, utengenezaji wa maudhui yanayobadilika, au sasisho za maendeleo zinazoendelea. Waendelezaji wanaojenga mawakala wa AI wa hali ya juu mara nyingi hukabiliwa na changamoto wakati mifumo ya kazi inahitaji kusitisha, kukusanya ufafanuzi, au kuripoti hali wakati wa operesheni ndefu. Hali isiyobadilika, ya njia moja ya utekelezaji usio na hali inazuia ukuzaji wa programu za AI shirikishi na zinazojibu kweli.

Sasa, Amazon Bedrock AgentCore Runtime inaleta uwezo wa mteja wa Itifaki ya Muktadha wa Modeli (MCP) yenye hali (stateful) unaovunja ardhi, ikibadilisha jinsi mawakala wa AI wanavyoshirikiana na watumiaji na modeli kubwa za lugha (LLMs). Sasisho hili muhimu linawakomboa mawakala kutoka kwa vikwazo vya mawasiliano yasiyo na hali, kuwezesha mifumo ya kazi changamano, ya hatua nyingi, na shirikishi sana. Kwa kuunganisha vipengele muhimu vya mteja wa MCP – Uchochezi, Sampuli, na Arifa za Maendeleo – Bedrock AgentCore Runtime inawezesha mazungumzo ya pande mbili kati ya seva za MCP na wateja, ikifungua njia kwa suluhisho za AI zenye akili zaidi, zinazomlenga mtumiaji.

Kutoka Isiyo na Hali (Stateless) hadi yenye Hali (Stateful): Kufungua Mifumo ya Kazi ya Wakala Shirikishi

Hapo awali, usaidizi wa seva ya MCP kwenye AgentCore ulifanya kazi katika hali isiyo na hali (stateless), ambapo kila ombi la HTTP lilifanya kazi kwa kujitegemea, bila muktadha wowote wa pamoja. Ingawa hii ilirahisisha usambazaji kwa seva za zana za msingi, ilizuia sana matukio yanayohitaji mwendelezo wa mazungumzo, ufafanuzi wa mtumiaji katikati ya mfumo wa kazi, au ripoti ya maendeleo ya wakati halisi. Seva haikuweza kudumisha nyuzi ya mazungumzo katika maombi tofauti, ikizuia ukuzaji wa mawakala shirikishi kweli.

Kuanzishwa kwa uwezo wa mteja wa MCP wenye hali (stateful) kunabadilisha kimsingi dhana hii. Kwa kuweka stateless_http=False wakati wa kuanzisha seva, Wakati wa Kuendesha wa AgentCore hutoa microVM maalum kwa kila kikao cha mtumiaji. MicroVM hii hudumu kwa muda wote wa kikao—hadi saa 8, au dakika 15 za kutokuwa na shughuli kulingana na mpangilio wa idleRuntimeSessionTimeout—kuhakikisha kutengwa kwa CPU, kumbukumbu, na mfumo wa faili kati ya vikao. Mwendelezo unadumishwa kupitia kichwa cha Mcp-Session-Id, ambacho seva hutoa wakati wa kuanzisha na mteja hujumuisha katika maombi yote yanayofuata ili kuelekeza kwenye kikao kilekile. Mazingira haya maalum, yanayodumu yanaruhusu mawakala kukumbuka muktadha, kuomba maoni ya mtumiaji, kutengeneza maudhui ya LLM yanayobadilika, na kutoa sasisho zinazoendelea.

Jedwali lifuatalo linafupisha tofauti kuu kati ya hali zisizo na hali (stateless) na zenye hali (stateful):

	Stateless mode	Stateful mode
`stateless_http` setting	`TRUE`	`FALSE`
Session isolation	Dedicated microVM per session	Dedicated microVM per session
Session lifetime	Up to 8 hours; 15-min idle timeout	Up to 8 hours; 15-min idle timeout
Client capabilities	Not supported	Elicitation, sampling, progress notifications
Recommended for	Simple tool serving	Interactive, multi-turn workflows

Kikao kinapoisha au seva inapoanzishwa upya, maombi yanayofuata yenye kitambulisho cha kikao cha awali yatarudisha 404. Katika hatua hiyo, wateja lazima waanzishe upya muunganisho ili kupata kitambulisho kipya cha kikao na kuanza kikao kipya. Mabadiliko ya usanidi ili kuwezesha hali yenye hali ni alama moja tu katika uanzishaji wa seva yako:

mcp.run( transport="streamable-http", host="0.0.0.0", port=8000, stateless_http=False # Enable stateful mode)

Zaidi ya alama hii, uwezo wa mteja tatu unapatikana kiotomatiki mara tu mteja wa MCP anapotangaza kuunga mkono wakati wa kuanzisha mawasiliano.

Kuzama Kina katika Uwezo Mpya wa Mteja: Uchochezi, Sampuli, na Maendeleo

Kwa mabadiliko kuelekea hali yenye hali (stateful mode), Wakati wa Kuendesha wa Amazon Bedrock AgentCore unafungua uwezo tatu wenye nguvu wa mteja kutoka kwa ufafanuzi wa MCP, kila mmoja umeundwa kushughulikia mifumo tofauti ya mwingiliano muhimu kwa mawakala wa hali ya juu wa AI. Uwezo huu unabadilisha kile ambacho hapo awali kilikuwa utekelezaji mgumu, wa amri ya njia moja kuwa mazungumzo ya pande mbili, ya maji kati ya seva ya MCP na wateja wake waliounganishwa. Ni muhimu kutambua kwamba vipengele hivi ni vya hiari, ikimaanisha wateja wanatangaza usaidizi wao wakati wa kuanzisha, na seva lazima zitumie tu uwezo ambao mteja aliyeunganishwa ameuonyesha.

Uchochezi: Kuwezesha Pembejeo ya Mtumiaji Inayobadilika katika Mawakala wa AI

Uchochezi unasimama kama nguzo kuu ya AI shirikishi, ikiruhusu seva ya MCP kusitisha utekelezaji wake kwa busara na kuomba pembejeo maalum, iliyopangwa kutoka kwa mtumiaji kupitia mteja. Uwezo huu unaipa zana uwezo wa kuuliza maswali sahihi kwa nyakati zinazofaa ndani ya mfumo wake wa kazi, iwe ni kuthibitisha uamuzi, kukusanya mapendeleo ya mtumiaji, au kukusanya thamani iliyotokana na operesheni zilizotangulia. Seva inaanzisha hili kwa kutuma ombi la elicitation/create JSON-RPC, ambalo linajumuisha ujumbe unaosomeka na binadamu na requestedSchema ya hiari inayofafanua muundo wa jibu unaotarajiwa.

Ufafanuzi wa MCP unatoa aina mbili thabiti za uchochezi:

Aina ya fomu: Hii ni bora kwa kukusanya data iliyopangwa moja kwa moja kupitia mteja wa MCP, kama vile vigezo vya usanidi, mapendeleo ya mtumiaji, au uthibitisho rahisi ambapo data nyeti haihusiki.
Aina ya URL: Kwa mwingiliano unaohitaji mchakato salama, nje ya bendi, kama vile mtiririko wa OAuth, usindikaji wa malipo, au uingizaji wa taarifa nyeti, aina ya URL inamwelekeza mtumiaji kwenye URL ya nje. Hii inahakikisha kwamba taarifa nyeti hazipiti kabisa mteja wa MCP, ikiboresha usalama na utii.

Baada ya kupokea ombi la uchochezi, mteja hutoa kiolesura kinachofaa cha kuingiza data. Kitendo kinachofuata cha mtumiaji huamsha mfumo wa majibu wa vitendo vitatu kwa seva: accept (mtumiaji alitoa data iliyoombwa), decline (mtumiaji alikataa waziwazi ombi), au cancel (mtumiaji aliondoa ujumbe bila kufanya chaguo). Seva zenye akili zimeundwa kushughulikia kila moja ya matukio haya kwa ustadi, kuhakikisha uzoefu thabiti na rafiki kwa mtumiaji. Kwa mfano, zana ya add_expense_interactive, kama inavyoonyeshwa katika nyenzo chanzo, inaweza kumwongoza mtumiaji kupitia mfululizo wa maswali—kiasi, maelezo, kategoria, na uthibitisho wa mwisho—kabla ya kuweka data kwenye backend kama Amazon DynamoDB. Kila hatua inatumia modeli za Pydantic kufafanua pembejeo inayotarajiwa, ambayo FastMCP hubadilisha bila mshono kuwa Schema ya JSON inayohitajika kwa ombi la elicitation/create.

Sampuli na Arifa za Maendeleo: Kuongeza Mwingiliano na Uwazi wa LLM

Zaidi ya mwingiliano wa moja kwa moja wa mtumiaji, Sampuli inaiwezesha seva ya MCP kuwa na uwezo wa kuomba maudhui yanayotokana na LLM moja kwa moja kutoka kwa mteja kupitia sampling/createMessage. Hii ni utaratibu muhimu kwani inaruhusu mantiki ya zana kwenye seva kutumia uwezo mkubwa wa modeli ya lugha bila kuhitaji kudhibiti sifa zake za LLM au miunganisho ya moja kwa moja ya API. Seva inatoa tu ujumbe na mapendeleo ya hiari ya modeli, na mteja, akifanya kazi kama mpatanishi, husambaza ombi kwa LLM yake iliyounganishwa na kurudisha jibu lililotokana. Hii inafungua matumizi mengi ya kivitendo, ikiwa ni pamoja na kuunda muhtasari wa kibinafsi, kutoa maelezo ya lugha asilia kutoka kwa data iliyopangwa, au kutoa mapendekezo yanayozingatia muktadha kulingana na mazungumzo yanayoendelea.

Kwa operesheni zinazochukua muda, Arifa za Maendeleo zinakuwa muhimu sana. Uwezo huu unaruhusu seva ya MCP kuripoti sasisho za nyongeza wakati wa kazi ndefu. Kwa kutumia ctx.report_progress(progress, total), seva inaweza kutoa sasisho zinazoendelea ambazo wateja wanaweza kuzitafsiri kuwa maoni yanayoonekana, kama vile upau wa maendeleo au kiashiria cha hali. Iwe ni kutafuta katika vyanzo vingi vya data au kutekeleza kazi changamano za kompyuta, sasisho za maendeleo zilizo wazi huhakikisha watumiaji wanabaki kufahamishwa, kuzuia kukasirika na kuboresha uzoefu wa jumla wa mtumiaji, badala ya kuwaacha wakitazama skrini tupu wakijiuliza kama mfumo bado unafanya kazi.

Kuandaa Maendeleo ya Wakala wa AI kwa Baadaye kwa Kutumia Bedrock AgentCore Runtime

Kuanzishwa kwa uwezo wa mteja wa MCP wenye hali (stateful) kwenye Wakati wa Kuendesha wa Amazon Bedrock AgentCore kunaashiria hatua kubwa mbele katika ukuzaji wa mawakala wa AI. Kwa kubadilisha mwingiliano uliokuwa hauna hali (stateless) hapo awali kuwa mazungumzo yenye nguvu, ya pande mbili, AWS inawawezesha waendelezaji kujenga programu za AI zenye akili zaidi, zinazojibu, na rafiki kwa mtumiaji. Uwezo huu – Uchochezi kwa pembejeo ya mtumiaji inayoongozwa, Sampuli kwa utengenezaji wa LLM wa mahitaji, na Arifa za Maendeleo kwa uwazi wa wakati halisi – kwa pamoja hufungua enzi mpya ya mifumo ya kazi ya wakala shirikishi. Kadiri AI inavyoendelea kubadilika, uwezo huu wa msingi utakuwa muhimu kwa kuunda AI wakala shirikishi ya hali ya juu ambayo inaweza kuunganishwa bila mshono katika michakato changamano ya biashara, kurekebisha mahitaji ya mtumiaji, na kutoa thamani ya kipekee.