This thesis proposed a framework in completing transactions among decentralized agents without the existence of trusted authority nor intermediate facilitator in facilitating the transactions. The framework consists of transaction concept and logic, set of algorithms to accomplish transactions, and architecture that enable agents to govern and self-organize accounting, authentication, and authorization of transactions for mutual benefit. The system of transaction is based on the concept of creation and elimination of institutional facts as the fundament to develop the transaction logic. Set of agents united in a trading community collectively accept set of arbitrary constructed assertion of facts establishing the presence of accounts of finances of transacting agents. During transactions, those accounts are altered through set of collaborative actions that agents perform specified by transaction algorithm. This social-based accounting mechanism introduces a mechanism that enables agents to collectively govern transactions in individual level. By considering reputation of the seller as well as the ratio between potential risk and benefit of every transaction, agents perform vote-based collective decision to authorize or not to authorize the transaction. The feedback of each transaction is made public altering reputation of the corresponding seller. The collective authorization mechanism establishes social control to block possible bad behavior, to protect buyer from risky transaction, and to induce good behavior. The result of simulation on this scheme shows how such social control works. It filters bad behaving agents and preserves good behaving agents. This condition is necessary to sustain trust within community and thus preserves the collaboration. A chapter of this thesis presents set of protocols developed from transaction algorithm that fulfills best-practice considerations. It successfully applies distributed cryptography to devise mutual authentication problem and to reduce communication cost as well as to eliminate scalability issue of original algorithm. Finally, an application of the framework is exemplified: an architecture of decentralized trading network that superimposes existing P2P networks allowing Internet users to trade for their mutual benefit.